1 2018-07-05 Martin Sebor <msebor@redhat.com>
4 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5 than its domain to compute its the upper bound of a char array.
7 2018-07-05 Nathan Sidwell <nathan@acm.org>
9 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
10 * doc/cpp.texi: Update comment.
11 * doc/tm.texi: Rebuilt.
12 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
14 * doc/extend.texi (Backwards Compatibility): Clarify it is system
15 headers affected by extern "C".
16 * system.h: Poison NO_IMPLICIT_EXTERN_C.
17 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
18 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
19 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
20 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
21 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
22 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
23 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
24 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
25 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
26 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
27 config/sparc/sp64-elf.h, config/spu/spu.h,
28 config/stormy16/stormy16.h, config/v850/v850.h,
29 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
30 define NO_IMPLICIT_EXTERN_C.
31 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
33 2018-07-05 Tamar Christina <tamar.christina@arm.com>
36 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
37 instead of GET_MODE_SIZE when comparing Units.
39 2018-07-05 Tamar Christina <tamar.christina@arm.com>
42 * rtlanal.c (set_noop_p): Constrain on mode change,
43 include hard-reg-set.h
45 2018-07-05 Tamar Christina <tamar.christina@arm.com>
47 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
49 2018-07-05 Jakub Jelinek <jakub@redhat.com>
52 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
55 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
57 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
59 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
62 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
64 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
66 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
69 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
71 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
73 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
76 2018-07-04 Martin Liska <mliska@suse.cz>
78 * tree-switch-conversion.c: Define
79 max_ratio_for_speed and max_ratio_for_size constants.
81 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
82 Martin Liska <mliska@suse.cz>
87 * common.opt: Rename align options with 'str_' prefix.
88 * common/config/i386/i386-common.c (set_malign_value): New
90 (ix86_handle_option): Use it to set -falign-* options/
91 * config/aarch64/aarch64-protos.h (struct tune_params): Change
92 type from int to string.
93 * config/aarch64/aarch64.c: Update default values from int
95 * config/alpha/alpha.c (alpha_override_options_after_change):
97 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
98 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
99 max skip conditionally.
100 * config/i386/freebsd.h (SUBALIGN_LOG): New.
101 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
102 max skip conditionally.
103 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
104 max skip conditionally.
105 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
106 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
107 max skip conditionally.
108 * config/i386/i386.c (struct ptt): Change type from int to
110 (ix86_default_align): Set default values.
111 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
112 max skip conditionally.
113 * config/i386/iamcu.h (SUBALIGN_LOG): New.
114 (ASM_OUTPUT_MAX_SKIP_ALIGN):
115 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
116 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
117 max skip conditionally.
118 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
119 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
120 * config/i386/x86-64.h (SUBALIGN_LOG): New.
121 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
122 max skip conditionally.
123 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
124 * config/ia64/ia64.c (ia64_option_override): Set default values
125 for alignment options.
126 * config/m68k/m68k.c: Handle new str_align_* options.
127 * config/mips/mips.c (mips_set_compression_mode): Change
129 (mips_option_override): Set default values for options.
130 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
132 * config/rs6000/rs6000.c (rs6000_option_override_internal):
134 * config/rx/rx.c (rx_option_override): Likewise.
135 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
136 (LABEL_ALIGN): Use align_labels_log.
137 (LOOP_ALIGN): Use align_loops_align.
138 * config/s390/s390.c (s390_asm_output_function_label): Use new
140 * config/sh/sh.c (sh_override_options_after_change):
141 Change type of constants.
142 * config/spu/spu.c (spu_sched_init): Likewise.
143 * config/sparc/sparc.c (sparc_option_override): Set default
145 * config/visium/visium.c (visium_option_override): Likewise.
146 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
147 emit p2align format with last argument if it's not needed.
148 * doc/invoke.texi: Document extended format of -falign-*.
149 * final.c: Use align_labels alignment.
150 * flags.h (struct target_flag_state): Change type to use
152 (struct align_flags_tuple): New.
153 (struct align_flags): Likewise.
154 (align_loops_log): Redefine macro to use new types.
155 (align_loops_max_skip): Redefine macro to use new types.
156 (align_jumps_log): Redefine macro to use new types.
157 (align_jumps_max_skip): Redefine macro to use new types.
158 (align_labels_log): Redefine macro to use new types.
159 (align_labels_max_skip): Redefine macro to use new types.
160 (align_functions_log): Redefine macro to use new types.
161 (align_loops): Redefine macro to use new types.
162 (align_jumps): Redefine macro to use new types.
163 (align_labels): Redefine macro to use new types.
164 (align_functions): Redefine macro to use new types.
165 (align_functions_max_skip): Redefine macro to use new types.
166 (align_loops_value): New macro.
167 (align_jumps_value): New macro.
168 (align_labels_value): New macro.
169 (align_functions_value): New macro.
170 * function.c (invoke_set_current_function_hook): Propagate
171 alignment values from flags to global variables default in
173 * ipa-icf.c (sem_function::equals_wpa): Use
174 cl_optimization_option_eq instead of memcmp.
175 * lto-streamer.h (cl_optimization_stream_out): Support streaming
177 (cl_optimization_stream_in): Likewise.
178 * optc-save-gen.awk: Support strings in cl_optimization.
179 * opth-gen.awk: Likewise.
180 * opts.c (finish_options): Remove error checking of invalid
182 (MAX_CODE_ALIGN): Remove.
183 (MAX_CODE_ALIGN_VALUE): Likewise.
184 (parse_and_check_align_values): New function.
185 (check_alignment_argument): Likewise.
186 (common_handle_option): Use check_alignment_argument.
187 * opts.h (parse_and_check_align_values): Declare.
188 * toplev.c (init_alignments): Remove.
189 (read_log_maxskip): New.
190 (parse_N_M): Likewise.
191 (parse_alignment_opts): Likewise.
192 (backend_init_target): Remove usage of init_alignments.
193 * toplev.h (parse_alignment_opts): Declare.
194 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
196 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
197 * tree.c (cl_option_hasher::equal): New.
198 * varasm.c: Use new global macros.
200 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
202 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
203 Use a simpler align directive also if MAXSKIP = ALIGN-1.
204 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
205 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
206 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
207 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
208 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
209 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
210 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
211 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
212 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
213 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
214 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
216 2018-07-04 Martin Liska <mliska@suse.cz>
217 Jonathan Wakely <jwakely@redhat.com>
219 * coverage.c: Use correct type.
220 * doc/invoke.texi: Language correction.
222 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
225 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
226 ENDBRANCH for non-tail call which may return via indirect branch.
227 * doc/extend.texi: Document indirect_return attribute.
229 2018-07-03 Martin Sebor <msebor@redhat.com>
231 PR tree-optimization/86274
232 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
234 (format_floating): Correct handling of infinities and NaNs.
236 2018-07-03 Martin Sebor <msebor@redhat.com>
238 * print-tree.c (print_real_cst): New function.
239 (print_node_brief): Call it.
242 2018-07-03 Jeff Law <law@redhat.com>
244 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
245 into a single pattern.
247 * config/h8300/h8300.md (ors code_iterator): New.
248 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
249 a single pattern and single splitter.
250 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
251 (iorqi3_1, xorqi3_1): Likewise.
252 (iorqi3, xorqi3 expanders): Similarly.
254 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
255 (movmd_internal) into a single pattern using the P mode iterator.
256 (movmd splitters): Similarly.
257 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
258 (movsd splitters): Similarly.
260 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
261 ADDB, ADDW and ADDL into a single ADD attribute which selects the
262 right table based on the size of the operand.
263 * config/h8300/h8300.md (length_table): Corresponding changes. All
264 references to "addb", "addw" and "addl" changed to "add".
265 (btst patterns): Merge two variants into a single pattern.
266 (tstqi, tsthi): Likewise.
267 (addhi3_incdec, addsi3_incdec): Likewise.
268 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
269 (mulhi3, mulsi3): Likewise.
270 (udivhi3, udivsi3): Likewise.
271 (divhi3, divsi3): Likewise.
272 (andorqi3, andorhi3, andorsi3): Likewise.
274 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
277 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
278 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
280 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
282 PR tree-optimization/85694
283 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
284 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
286 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
287 (<u>avg<mode>3_ceil): New patterns.
289 2018-07-03 David Malcolm <dmalcolm@redhat.com>
291 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
292 scan-tree-dump directive.
293 * gcc.dg/vect/slp-perm-2.c: Likewise.
294 * gcc.dg/vect/slp-perm-3.c: Likewise.
295 * gcc.dg/vect/slp-perm-5.c: Likewise.
296 * gcc.dg/vect/slp-perm-6.c: Likewise.
297 * gcc.dg/vect/slp-perm-7.c: Likewise.
298 * gcc.dg/vect/slp-perm-8.c: Likewise.
300 2018-07-03 Marek Polacek <polacek@redhat.com>
303 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
305 2018-07-03 Richard Biener <rguenther@suse.de>
308 * tree-ssa-structalias.c (find_func_clobbers): Properly
309 handle indirect calls.
311 2018-07-03 Jeff Law <law@redhat.com>
313 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
314 (shifts): New code iterator.
315 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
316 expander. Fix HImode handling on H8/SX.
317 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
318 (subqi3, subhi3, subsi3 expanders): Likewise.
319 (andqi3, andhi3, andsi3 expanders): Likewise.
320 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
321 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
322 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
323 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
324 (zero_extendqihi2, zero_extendqisi2): Likewise.
325 (extendqihi2, extendqisi2): Likewise.
326 (rotlqi3, rotlhi3, rotlsi3): Likewise.
327 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
328 (rotlqi3_1, rotlhi3_1): Likewise.
329 (logicalhi3_sn, logicalsi3_sn): Likewise.
330 (logicalhi3, logicalsi3): Likewise.
332 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
334 * tree-vect-patterns.c (vect_recog_rotate_pattern)
335 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
336 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
337 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
338 type to append_pattern_def_seq instead of creating a stmt_vec_info
340 (build_mask_conversion): Likewise. Remove vinfo argument.
341 (vect_add_conversion_to_patterm): Likewise, renaming to...
342 (vect_add_conversion_to_pattern): ...this.
343 (vect_recog_mask_conversion_pattern): Update call to
344 build_mask_conversion. Pass the vector type to
345 append_pattern_def_seq here too.
346 (vect_recog_gather_scatter_pattern): Update call to
347 vect_add_conversion_to_pattern.
349 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
351 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
352 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
353 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
354 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
355 STMT_VINFO_PATTERN_DEF_SEQ to null here.
356 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
357 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
358 append_pattern_def_seq instead of new_pattern_def_seq.
359 (vect_recog_divmod_pattern): Do both of the above.
360 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
363 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
365 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
366 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
367 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
368 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
369 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
370 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
371 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
372 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
373 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
374 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
375 parameter with a single stmt_vec_info.
376 (vect_recog_func_ptr): Likewise.
377 (vect_recog_gather_scatter_pattern): Likewise, folding in...
378 (vect_try_gather_scatter_pattern): ...this.
379 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
380 the stmt_vec_info of the statement to be matched. Don't clear
381 STMT_VINFO_RELATED_STMT.
382 (vect_pattern_recog): Update call accordingly.
384 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
386 PR tree-optimization/85694
387 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
388 (uavgM3_ceil): Document new optabs.
389 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
390 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
392 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
393 (savg_ceil_optab): New optabs.
394 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
395 (vect_vect_recog_func_ptrs): Add it.
396 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
397 constant directly from the associated lhs.
399 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
401 * tree-vect-patterns.c (vect_split_statement): New function.
402 (vect_convert_input): Use it to try to split an existing cast.
404 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
406 * poly-int.h (print_hex): New function.
407 * dumpfile.h (dump_dec, dump_hex): Declare.
408 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
409 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
410 min_input_precision, operation_precision and operation_sign.
411 * tree-vect-patterns.c (vect_get_range_info): New function.
412 (vect_same_loop_or_bb_p, vect_single_imm_use)
413 (vect_operation_fits_smaller_type): Delete.
414 (vect_look_through_possible_promotion): Add an optional
415 single_use_p parameter.
416 (vect_recog_over_widening_pattern): Rewrite to use new
417 stmt_vec_info infomration. Handle one operation at a time.
418 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
419 (vect_truncatable_operation_p, vect_set_operation_type)
420 (vect_set_min_input_precision): New functions.
421 (vect_determine_min_output_precision_1): Likewise.
422 (vect_determine_min_output_precision): Likewise.
423 (vect_determine_precisions_from_range): Likewise.
424 (vect_determine_precisions_from_users): Likewise.
425 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
426 (vect_vect_recog_func_ptrs): Put over_widening first.
428 (vect_pattern_recog): Call vect_determine_precisions.
430 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
432 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
433 statements that have been replaced by further pattern statements.
434 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
436 2018-07-03 Richard Biener <rguenther@suse.de>
438 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
439 always set *dt. Dump vectype in vectype overload.
440 * dumpfile.h (dump_gimple_expr): New function.
441 (dump_gimple_expr_loc): Likewise.
442 * dumpfile.c (dump_gimple_expr): New function.
443 (dump_gimple_expr_loc): Likewise.
445 2018-07-02 Jeff Law <law@redhat.com>
447 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
448 the H8/300, H8/300H and H8/S variants into a single pattern.
449 (movhi_h8300, movqi_h8300hs): Similarly.
450 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
451 (QHI mode iterator): New.
453 * config/h8300/h8300.md: Remove trailing whitespace.
455 2018-07-02 Jim Wilson <jimw@sifive.com>
457 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
458 instead of emit_insn for interrupt returns.
459 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
460 (riscv_sret, riscv_uret): Likewise.
462 2018-07-02 David Malcolm <dmalcolm@redhat.com>
464 * pretty-print.c (selftest::test_pp_format): Move save and restore
465 of quotes to class auto_fix_quotes, and add an instance.
466 * selftest.c: Include "intl.h".
467 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
468 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
469 * selftest.h (selftest::auto_fix_quotes): New class.
471 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
473 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
474 (aarch64_sve_prepare_conditional_op): Remove.
475 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
476 Allow aarch64_simd_reg_or_zero as select operand; remove
477 the aarch64_sve_prepare_conditional_op call.
478 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
479 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
480 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
481 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
482 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
483 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
484 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
485 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
486 and a splitters to match all of the *_any patterns.
487 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
489 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
490 (SVE_COND_FP_BINARY_REV): Remove.
491 (sve_int_op_rev, sve_fp_op_rev): New.
492 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
493 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
494 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
495 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
496 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
497 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
498 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
499 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
500 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
502 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
503 Remove match_dup 1 from the inner unspec.
504 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
506 * config/aarch64/aarch64.md (movprfx): New attr.
507 (length): Default movprfx to 8.
508 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
509 (*madd<SVE_I>, *msub<SVE_I): Likewise.
510 (*<su>mul<SVE_I>3_highpart): Likewise.
511 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
512 (*v<ASHIFT><SVE_I>3): Likewise.
513 (*<su><MAXMIN><SVE_I>3): Likewise.
514 (*<su><MAXMIN><SVE_F>3): Likewise.
515 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
516 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
517 (*div<SVE_F>4): Likewise.
519 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
521 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
524 2018-07-02 Richard Biener <rguenther@suse.de>
526 PR tree-optimization/86363
527 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
528 memset argument refers to a non-variable address.
530 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
532 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
533 {PLUS,MINUS}_EXPR code to...
534 (adjust_symbolic_bound): ...here,
535 (combine_bound): ...here,
536 (set_value_range_with_overflow): ...and here.
538 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
540 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
542 (extract_range_from_abs_expr): ...here.
544 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
546 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
547 -fno-omit-frame-pointer when not optimizing.
549 2018-07-02 Martin Liska <mliska@suse.cz>
552 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
553 (propagate_nothrow): Likewise.
555 2018-07-02 Martin Liska <mliska@suse.cz>
558 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
560 2018-07-02 David Malcolm <dmalcolm@redhat.com>
562 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
563 function in r262149, changing "loc" param from source_location to
564 const dump_location_t &.
565 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
566 declaration, as above.
568 2018-07-01 Paul Koning <ni1d@arrl.net>
570 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
571 -munit-asm, -mgnu-asm, -mdec-asm.
572 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
573 (pdp11_output_labelref): New.
574 (pdp11_output_def): New.
575 (pdp11_output_addr_vec_elt): New.
576 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
577 %# and %@ format codes.
578 (pdp11_option_override): New.
579 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
580 (pdp11_output_ident): New.
581 (pdp11_asm_named_section): New.
582 (pdp11_asm_init_sections): New.
583 (pdp11_file_start): New.
584 (pdp11_file_end): New.
585 (output_ascii): Use .ascii/.asciz for -mdec-asm.
586 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
587 %o, like %c but octal.
588 (pdp11_option_override): New.
589 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
591 (DATA_SECTION_ASM_OP): Ditto.
592 (READONLY_DATA_SECTION_ASM_OP): New.
593 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
594 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
595 (ASM_OUTPUT_LABELREF): Ditto.
596 (ASM_OUTPUT_DEF): Ditto.
597 (ASM_OUTPUT_EXTERNAL): New.
598 (ASM_OUTPUT_SOURCE_FILENAME): New.
599 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
600 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
601 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
602 %# and %@ format codes.
603 * config/pdp11/pdp11.opt (mgnu-asm): New.
604 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
605 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
606 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
608 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
610 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
611 dereferencing path[] beyond its length.
613 (debug_all_paths): New.
614 (rewire_first_differing_edge): New.
615 (adjust_paths_after_duplication): New.
616 (duplicate_thread_path): Call adjust_paths_after_duplication.
618 (thread_through_all_blocks): Add new argument to
619 duplicate_thread_path.
621 2018-06-30 Jim Wilson <jimw@sifive.com>
623 * config/riscv/predicates.md (p2m1_shift_operand): New.
624 (high_mask_shift_operand): New.
625 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
626 pattern using p2m1_shift_operand.
627 (lshsi3_zero_extend_3+2): New combiner pattern using
628 high_mask_shift_operand.
630 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
632 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
634 (vect_recog_rotate_pattern): ...here.
635 (vect_convert_input): Try to insert casts of invariants in the
637 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
638 preheader to be empty.
640 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
642 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
643 vector type. If given, install it in the new statement's
645 (vect_element_precision): New function.
646 (vect_unpromoted_value): New struct.
647 (vect_unpromoted_value::vect_unpromoted_value): New function.
648 (vect_unpromoted_value::set_op): Likewise.
649 (vect_look_through_possible_promotion): Likewise.
650 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
651 (vect_widened_op_tree, vect_convert_input): Likewise.
652 (vect_convert_inputs, vect_convert_output): Likewise.
653 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
654 to handle the optional cast of the multiplication result and
655 vect_widened_op_tree to detect the widened multiplication itself.
656 Do not require the input and output of promotion casts to have
657 the same sign, but base the signedness of the operation on the
658 input rather than the result. If the pattern includes two
659 promotions, check that those promotions have the same sign.
660 Do not restrict the MULT_EXPR handling to a double-width result;
661 handle quadruple-width results and wider. Use vect_convert_inputs
662 to convert the inputs to the common type.
663 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
664 to handle the optional cast of the ABS result. Also allow a sign
665 change or a sign extension between the ABS and MINUS.
666 Use vect_widened_op_tree to detect the widened subtraction and use
667 vect_convert_inputs to convert the inputs to the common type.
668 (vect_handle_widen_op_by_const): Delete.
669 (vect_recog_widen_op_pattern): New function.
670 (vect_recog_widen_mult_pattern): Use it.
671 (vect_recog_widen_shift_pattern): Likewise.
672 (vect_recog_widen_sum_pattern): Use
673 vect_look_through_possible_promotion to handle the promoted
676 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
678 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
679 the containing gimple_seq *.
680 * gimple-iterator.h (gsi_for_stmt): Declare it.
681 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
682 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
683 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
684 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
685 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
687 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
689 (vect_mark_pattern_stmts): ...here. Handle cases in which the
690 statement being replaced is part of an existing pattern
691 definition sequence, inserting the new pattern statements before
693 (vect_pattern_recog_1): Don't return a bool. If the statement
694 is already part of a pattern, instead apply pattern matching
695 to the pattern definition statements. Don't clear the
696 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
697 (vect_pattern_recog): Don't break after the first match;
698 continue processing the pattern definition statements instead.
699 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
701 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
703 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
704 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
705 (vect_recog_widen_sum_pattern): Use it.
707 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
709 * tree-vect-loop.c (vectorizable_reduction): Assert that the
710 phi is not a pattern statement and has not been replaced by
712 * tree-vect-patterns.c (type_conversion_p): Don't check
713 STMT_VINFO_IN_PATTERN_P.
714 (vect_recog_vector_vector_shift_pattern): Likewise.
715 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
716 the pattern statement rather than the original statement; check
717 directly for a WIDEN_MULT_EXPR here.
718 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
719 vect_is_simple_use to return the pattern statement rather
720 than the original statement; use is_pattern_stmt_p to check
721 for such a pattern statement.
722 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
723 to return the pattern statement rather than the original statement;
724 don't do the same transformation here.
725 (vect_is_simple_use): If the defining statement has been replaced
726 by a pattern statement, return the pattern statement instead.
727 Remove the corresponding (local) transformation from the vectype
730 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
732 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
733 end and default to null.
734 * tree-vect-loop.c (vect_create_epilog_for_reduction)
735 (vectorizable_reduction): Update calls accordingly, dropping the
736 gimple ** argument if the passed-back statement isn't needed.
737 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
738 (vect_recog_rotate_pattern): Likewise.
739 (vect_recog_mask_conversion_pattern): Likewise.
740 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
741 (vect_mask_constant_operand_p): Likewise.
742 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
743 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
744 (get_group_load_store_type, get_load_store_type): Likewise.
745 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
746 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
747 (vectorizable_conversion, vectorizable_assignment): Likewise.
748 (vectorizable_shift, vectorizable_operation): Likewise.
749 (vectorizable_store, vect_is_simple_cond): Likewise.
750 (vectorizable_condition, vectorizable_comparison): Likewise.
751 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
752 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
753 and move it to the end. Cope with null def_stmt_outs.
755 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
757 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
759 2018-06-29 Jeff Law <law@redhat.com>
761 * config/v850/v850.c (v850_legitimate_address_p): Handle large
762 displacements for TARGET_V850E2V3 and newer.
763 (TARGET_LRA_P): Remove. Defaults to LRA now.
764 * config/v850/v850.md (sign23byte_load): Remove.
765 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
766 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
768 2018-06-29 Martin Liska <mliska@suse.cz>
771 * coverage.c (coverage_init): Mangle full path name.
772 * doc/invoke.texi: Document the change.
773 * gcov-io.c (mangle_path): New.
774 * gcov-io.h (mangle_path): Likewise.
775 * gcov.c (mangle_name): Use mangle_path for path mangling.
777 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
779 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
780 if starting source register is not even.
782 2018-06-29 Martin Liska <mliska@suse.cz>
784 PR tree-optimization/86263
785 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
786 Make edge redirection.
788 2018-06-29 David Malcolm <dmalcolm@redhat.com>
790 * dumpfile.c (dump_loc): Add indentation based on scope depth.
791 (dump_scope_depth): New variable.
792 (get_dump_scope_depth): New function.
793 (dump_begin_scope): New function.
794 (dump_end_scope): New function.
795 * dumpfile.h (get_dump_scope_depth): New declaration.
796 (dump_begin_scope): New declaration.
797 (dump_end_scope): New declaration.
798 (class auto_dump_scope): New class.
799 (AUTO_DUMP_SCOPE): New macro.
800 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
803 2018-06-29 Richard Biener <rguenther@suse.de>
805 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
806 compute_all_dependences succeeds.
807 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
808 exceed --param loop-max-datarefs-for-datadeps.
810 2018-06-29 Jakub Jelinek <jakub@redhat.com>
812 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
814 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
817 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
818 alternative 0 in preferred_for_speed attribute.
820 2018-06-28 Paul Koning <ni1d@arrl.net>
822 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
823 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
824 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
825 * config/pdp11/pdp11.md: Correct "length" attribute calculation
826 for shift insn patterns.
828 2018-06-28 David Malcolm <dmalcolm@redhat.com>
830 * cgraph.c (cgraph_node::get_body): Replace assignments to
831 "dump_file" with calls to set_dump_file.
832 * dumpfile.c (alt_dump_file): Make static, and group with...
833 (alt_flags): ...this definition.
834 (dumps_are_enabled): New variable.
835 (refresh_dumps_are_enabled): New function.
836 (set_dump_file): New function.
837 (set_alt_dump_file): New function.
838 (gcc::dump_manager::dump_start): Replace assignments to
839 "dump_file" and "alt_dump_file" with calls to set_dump_file and
841 (gcc::dump_manager::dump_finish): Likewise.
842 * dumpfile.h (alt_dump_file): Delete decl.
843 (dumps_are_enabled): New variable decl.
844 (set_dump_file): New function decl.
845 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
847 * tree-nested.c (lower_nested_functions): Replace assignments to
848 "dump_file" with calls to set_dump_file.
850 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
852 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
853 goto_locus of each outgoing edge of each basic block.
855 2018-06-28 Richard Biener <rguenther@suse.de>
857 * dwarf2out.c (decl_scope_table): Remove.
858 (push_decl_scope): Likewise.
859 (pop_decl_scope): Likewise.
860 (gen_type_die_for_member): Do not call push/pop_decl_scope.
861 (gen_struct_or_union_type_die): Likewise.
862 (gen_tagged_type_die): Likewise.
863 (dwarf2out_init): Do not initialize decl_scope_table.
864 (dwarf2out_c_finalize): Do not free it.
866 2018-06-28 Richard Biener <rguenther@suse.de>
868 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
869 deciding whether to not re-use a DIE.
871 2018-06-28 Richard Biener <rguenther@suse.de>
873 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
874 DW_AT_abstract_origin attribute.
876 2018-06-28 Martin Liska <mliska@suse.cz>
878 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
879 Use newly introduced constants.
880 * tree-switch-conversion.h (struct jump_table_cluster):
881 Define max_ratio_for_size and max_ratio_for_speed.
883 2018-06-28 Martin Liska <mliska@suse.cz>
885 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
886 Add new checking assert to catch invalid state.
887 (jump_table_cluster::can_be_handled): Handle single case
889 (jump_table_cluster::is_beneficial): Bail out for such case.
890 (bit_test_cluster::find_bit_tests):
891 Add new checking assert to catch invalid state.
892 (bit_test_cluster::can_be_handled): Handle single case
894 (bit_test_cluster::is_beneficial): Bail out for such case.
895 (switch_decision_tree::analyze_switch_statement):
898 2018-06-28 Martin Liska <mliska@suse.cz>
900 * common.opt: Introduce -completion option.
901 * gcc.c (driver_handle_option): Handle it.
902 (driver::main): Print completions if completion
904 * opt-suggestions.c (option_proposer::get_completions):
906 (option_proposer::suggest_completion): Likewise.
907 (option_proposer::find_param_completions): Likewise.
908 (verify_autocompletions): Likewise.
909 (test_completion_valid_options): Likewise.
910 (test_completion_valid_params): Likewise.
911 (in_completion_p): Likewise.
912 (empty_completion_p): Likewise.
913 (test_completion_partial_match): Likewise.
914 (test_completion_garbage): Likewise.
915 (opt_proposer_c_tests): Likewise.
916 * opt-suggestions.h: Declare new functions.
917 * opts.c (common_handle_option): Handle OPT__completion_.
918 * selftest-run-tests.c (selftest::run_tests): Add
919 opt_proposer_c_tests.
920 * selftest.c (assert_str_startswith): New.
921 * selftest.h (assert_str_startswith): Likewise.
922 (opt_proposer_c_tests): New.
923 (ASSERT_STR_STARTSWITH): Likewise.
925 2018-06-28 Martin Liska <mliska@suse.cz>
927 * Makefile.in: Add opt-suggestions.o.
928 * gcc-main.c: Include opt-suggestions.h.
929 * gcc.c (driver::driver): Likewise.
930 (driver::~driver): Remove m_option_suggestions.
931 (driver::build_option_suggestions): Moved to option_proposer.
932 (driver::suggest_option): Likewise.
933 (driver::handle_unrecognized_options): Use option_proposer.
934 * gcc.h (class driver): Add new memver m_option_proposer.
935 * opt-suggestions.c: New file.
936 * opt-suggestions.h: New file.
938 2018-06-28 Martin Liska <mliska@suse.cz>
940 * vec.h (class auto_string_vec): New (moved from auto_argvec).
941 (auto_string_vec::~auto_string_vec): Likewise.
943 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
945 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
946 prevent_decl_creation_for_types fields up and add reset_location field.
947 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
948 statement if id->reset_location is true.
949 (copy_edges_for_bb): Do not set goto_locus on the new edges if
950 id->reset_location is true.
951 (copy_phis_for_bb): Force input_location on the arguments if
952 id->reset_location is true.
953 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
954 is set on the function to be inlined.
956 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
958 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
960 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
962 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
964 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
965 hard registers for the clobbered pseudo.
967 2018-06-27 Paul Koning <ni1d@arrl.net>
969 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
970 mutually exclusive options.
971 * config/pdp11/constraints.md (h): New constraint.
972 (O): Update definition to match shift code generation.
974 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
976 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
978 (output_jump): Change arguments.
979 (pdp11_fixed_cc_regs): New function.
980 (pdp11_cc_mode): Ditto.
981 (pdp11_expand_shift): Ditto.
982 (pdp11_assemble_shift): Ditto.
983 (pdp11_small_shift): Ditto.
984 (pdp11_branch_cost): Remove.
985 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
987 (pdp11_register_move_cost): Update for CC registers.
988 (pdp11_rtx_costs): Add case for LSHIFTRT.
989 (pdp11_output_jump): Add CCNZ mode conditional branches.
990 (notice_update_cc_on_set): Remove.
991 (pdp11_cc_mode): New function.
992 (simple_memory_operand): Correct pre/post decrement case.
993 (no_side_effect_operand): New function.
994 (pdp11_regno_reg_class): Add CC_REGS class.
995 (pdp11_fixed_cc_regs): New function.
996 (pdp11_small_shift): New function.
997 (pdp11_expand_shift): New function to expand shift insns.
998 (pdp11_assemble_shift): New function to output shifts.
999 (pdp11_branch_cost): Remove.
1000 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1001 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1002 (WCHAR_TYPE): Ditto.
1003 (PTRDIFF_TYPE): Ditto.
1004 (ADJUST_INSN_LENGTH): New macro.
1005 (FIXED_REGISTERS): Add CC registers.
1006 (CALL_USED_REGISTERS): Ditto.
1008 (REG_CLASS_NAMES): Ditto.
1009 (REG_CLASS_CONTENTS): Ditto.
1010 (SELECT_CC_MODE): Use new function.
1011 (TARGET_FLAGS_REGNUM): New macro.
1012 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1013 (cc0_reg_rtx): Remove.
1014 (CC_STATUS_MDEP): Remove.
1015 (CC_STATUS_MDEFP_INIT): Remove.
1016 (CC_IN_FPU): Remove.
1017 (NOTICE_UPDATE_CC): Remove.
1018 (REGISTER_NAMES): Add CC registers.
1019 (BRANCH_COST): Change to constant 1.
1020 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1022 * config/pdp11/pdp11.opt (mbcopy): Remove.
1023 (mbcopy-builtin): Remove.
1024 (mbranch-cheap): Remove.
1025 (mbranch-expensive): Remove.
1026 * config/pdp11/predicates.md (expand_shift_operand): Update to
1027 match shift code generation.
1028 (ccnz_operator): New predicate.
1029 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1030 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1031 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1033 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1034 description of O constraint.
1036 2018-06-27 Jeff Law <law@redhat.com>
1037 Austin Law <austinklaw@gmail.com>
1039 * config/v850/v850.md (addsi3_set_flags): New pattern.
1040 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1041 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1042 (zero_extendhisi2_v850_set_flags): Likewise.
1043 (zero_extendqisi2_v850_set_flags): Likewise.
1044 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1045 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1046 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1048 * config/v850/v850-protos.h (notice_update_cc): Remove.
1049 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1050 (v850_print_operand): Handle 'D' and "d".
1051 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1052 Add handling of arithmetic/logical operations compared against zero.
1053 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1054 Do not look at v850_compare_op, instead get mode from last argument.
1055 (v850_gen_compare): Remove
1056 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1057 after reload for prologue insns.
1058 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1060 (construct_save_jarl): Likewise.
1061 (TARGET_FLAGS_REGNUM): Define.
1062 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
1063 (NOTICE_UPDATE_CC): Remove.
1064 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
1065 than cc0. Conditionalize on reload_completed.
1066 (cmpsi_insn, setfcc_insn): Likewise.
1067 (tst1 splitter): Turn into define_and_split which sets the flags
1069 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
1070 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
1071 (cstoresf4, cstoredf4): Clobber the flags.
1072 (cmpsi, cmpsf, cmpdf): Remove expanders.
1073 (setf_insn): Remove pattern.
1074 (addsi3): Turn into define_and_split which clobbers the flags after
1075 reload and a suitable pattern (addsi3_clobber_flags) for use after
1077 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
1078 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
1079 (ashrsi3, ashrsi3_v850e2): Likewise.
1080 (bins): Clobber the flags.
1081 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
1082 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
1083 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
1084 (call_value_internal_short, call_value_internal_long): Likewise.
1085 (callt_save_interrupt, callt_return_interrupt): Likewise.
1086 (save_interrupt, return_interrupt): Likewise.
1087 (callt_save_all_interrupt, save_all_interrupt): Likewise.
1088 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
1089 (restore_all_interrupt, _restore_all_interrupt): Likewise.
1090 (All FP comparisons): Only allow after reload has completed.
1092 (divh, divhu): Tweak output template.
1093 (branch_z_normal, branch_z_invert): Remove
1094 (branch_nz_normal, branch_nz_invert): Likewise.
1095 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
1097 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
1098 * config/v850/v850.c (notice_update_cc): Remove.
1099 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
1100 (CC_NO_CARRY): Likewise.
1101 (NOTICE_UPDATE_CC): Define to nothing.
1102 * config/v850/v850.md: Remove block comment on cc0 handling
1103 Remove "cc" attribute from all patterns. Remove cc_status handling
1104 from all patterns. Minor formatting fixes.
1106 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1108 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
1109 (cortex-a76.cortex-a55): Likewise.
1110 * config/aarch64/aarch64-tune.md: Regenerate.
1111 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
1112 cortex-a76.cortex-a55.
1114 2018-06-27 Jeff Law <law@redhat.com>
1116 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
1117 (MULTILIB_DIRNAMES): Similarly.
1119 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
1121 * gimple.h (gimple_return_retbnd): Delete.
1122 (gimple_return_set_retbnd): Likewise.
1123 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
1124 gimple_return_set_retbnd.
1125 * gimple-pretty-print.c (dump_gimple_return): Remove call to
1126 gimple_return_retbnd and adjust.
1127 * tree-inline.h (struct copy_body_data): Remove retbnd field.
1128 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
1129 Explicitly return NULL in a couple more cases. Move assertion
1130 on debug statements and remove unreachable code.
1131 (reset_debug_binding): Do not test id->retbnd.
1132 (expand_call_inline): Do not set it.
1134 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1136 * configure.ac: Add --disable-gcov option.
1137 * configure: Regenerate.
1138 * Makefile.in: Honour @enable_gcov@.
1139 * doc/install.texi: Document --disable-gcov.
1141 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1143 * config/arm/arm-cpus.in (cortex-a76): New entry.
1144 (cortex-a76.cortex-a55): Likewise.
1145 * config/arm/arm-tables.opt: Regenerate.
1146 * config/arm/arm-tune.md: Likewise.
1147 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
1148 * doc/invoke.texi (ARM Options): Document cortex-a76 and
1149 cortex-a76.cortex-a55.
1151 2018-06-27 Tamar Christina <tamar.christina@arm.com>
1154 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
1156 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
1158 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
1160 (EPILOGUE_USES): Likewise.
1162 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
1164 * tree-inline.c (remap_location): New function extracted from...
1165 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
1166 (copy_phis_for_bb): ...here. Call remap_location.
1167 (copy_cfg_body): Adjust call to copy_edges_for_bb.
1169 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1171 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
1172 unaligned vsx for 16B memset.
1174 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
1177 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
1178 ieee128_float_type_node to long_double_type_node unless
1179 TARGET_LONG_DOUBLE_128 is set.
1181 2018-06-26 David Malcolm <dmalcolm@redhat.com>
1183 * cfgloop.c (get_loop_location): Convert return type from
1184 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
1185 by implicit construction from rtx_insn *, and using
1186 dump_user_location_t::from_function_decl for the fallback case.
1187 * cfgloop.h (get_loop_location): Convert return type from
1188 location_t to dump_user_location_t.
1189 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
1190 dump_printf_loc to pass in a dump_location_t rather than a
1191 location_t, via the gimple stmt.
1192 * coverage.c (get_coverage_counts): Update calls to
1193 dump_printf_loc to pass in dump_location_t rather than a
1195 * doc/optinfo.texi (Dump types): Convert example of
1196 dump_printf_loc from taking "locus" to taking "insn". Update
1197 description of the "_loc" calls to cover dump_location_t.
1198 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
1200 (dump_user_location_t::dump_user_location_t): New constructors,
1201 from gimple *stmt and rtx_insn *.
1202 (dump_user_location_t::from_function_decl): New function.
1203 (dump_loc): Make static.
1204 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
1205 const dump_location_t &.
1206 (dump_generic_expr_loc): Delete.
1207 (dump_printf_loc): Convert param "loc" from location_t to
1208 const dump_location_t &.
1209 (selftest::test_impl_location): New function.
1210 (selftest::dumpfile_c_tests): New function.
1211 * dumpfile.h: Include "profile-count.h".
1212 (class dump_user_location_t): New class.
1213 (struct dump_impl_location_t): New struct.
1214 (class dump_location_t): New class.
1215 (dump_printf_loc): Convert 2nd param from source_location to
1216 const dump_location_t &.
1217 (dump_generic_expr_loc): Delete.
1218 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
1219 const dump_location_t &.
1220 * gimple-fold.c (fold_gimple_assign): Update call to
1221 dump_printf_loc to pass in a dump_location_t rather than a
1222 location_t, via the gimple stmt.
1223 (gimple_fold_call): Likewise.
1224 * gimple-loop-interchange.cc
1225 (loop_cand::analyze_iloop_reduction_var): Update for change to
1226 check_reduction_path.
1227 (tree_loop_interchange::interchange): Update for change to
1229 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
1230 change in return-type of find_loop_location.
1231 (graphite_regenerate_ast_isl): Likewise.
1232 * graphite-optimize-isl.c (optimize_isl): Likewise.
1233 * graphite.c (graphite_transform_loops): Likewise.
1234 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
1235 pass in a dump_location_t rather than a location_t, via the
1237 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1238 * ipa.c (walk_polymorphic_call_targets): Likewise.
1239 * loop-unroll.c (report_unroll): Convert "locus" param from
1240 location_t to dump_location_t.
1241 (decide_unrolling): Update for change to get_loop_location's
1243 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
1244 location_t to dump_user_location_t.
1245 (grid_find_single_omp_among_assignments_1): Updates calls to
1246 dump_printf_loc to pass in a dump_location_t rather than a
1247 location_t, via the gimple stmt.
1248 (grid_parallel_clauses_gridifiable): Convert "tloc" from
1249 location_t to dump_location_t. Updates calls to dump_printf_loc
1250 to pass in a dump_location_t rather than a location_t, via the
1252 (grid_inner_loop_gridifiable_p): Likewise.
1253 (grid_dist_follows_simple_pattern): Likewise.
1254 (grid_gfor_follows_tiling_pattern): Likewise.
1255 (grid_target_follows_gridifiable_pattern): Likewise.
1256 (grid_attempt_target_gridification): Convert initialization
1257 of local "grid" from memset to zero-initialization; FIXME: does
1258 this require C++11? Update call to dump_printf_loc to pass in a
1259 optinfo_location rather than a location_t, via the gimple stmt.
1260 * profile.c (read_profile_edge_counts): Updates call to
1261 dump_printf_loc to pass in a dump_location_t rather than a
1263 (compute_branch_probabilities): Likewise.
1264 * selftest-run-tests.c (selftest::run_tests): Call
1266 * selftest.h (dumpfile_c_tests): New decl.
1267 * tree-loop-distribution.c (pass_loop_distribution::execute):
1268 Update for change in return type of find_loop_location.
1269 * tree-parloops.c (parallelize_loops): Likewise.
1270 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
1271 "locus" from location_t to dump_user_location_t.
1272 (canonicalize_loop_induction_variables): Likewise.
1273 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
1274 for change in return type of find_loop_location.
1275 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
1276 to dump_printf_loc to pass in a dump_location_t rather than a
1277 location_t, via the stmt.
1278 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
1280 * tree-vect-loop-manip.c (find_loop_location): Convert return
1281 type from source_location to dump_user_location_t.
1282 (vect_do_peeling): Update for above change.
1283 (vect_loop_versioning): Update for change in type of
1285 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
1286 from location_t to dump_user_location_t.
1287 (vect_estimate_min_profitable_iters): Update for change in type
1289 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
1290 location_t to dump_location_t.
1291 (vect_slp_bb): Update for change in type of vect_location.
1292 * tree-vectorizer.c (vect_location): Convert from source_location
1293 to dump_user_location_t.
1294 (try_vectorize_loop_1): Update for change in vect_location's type.
1295 (vectorize_loops): Likewise.
1296 (increase_alignment): Likewise.
1297 * tree-vectorizer.h (vect_location): Convert from source_location
1298 to dump_user_location_t.
1299 (find_loop_location): Convert return type from source_location to
1300 dump_user_location_t.
1301 (check_reduction_path): Convert 1st param from location_t to
1302 dump_user_location_t.
1303 * value-prof.c (check_counter): Update call to dump_printf_loc to
1304 pass in a dump_user_location_t rather than a location_t; update
1305 call to error_at for change in type of "locus".
1306 (check_ic_target): Update call to dump_printf_loc to
1307 pass in a dump_user_location_t rather than a location_t, via the
1310 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
1312 * config/s390/s390.h (enum processor_flags): Do not use
1313 default tune parameter when -march was specified.
1315 2018-06-26 Jakub Jelinek <jakub@redhat.com>
1318 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
1319 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
1322 2018-06-26 Richard Biener <rguenther@suse.de>
1324 PR tree-optimization/86287
1326 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
1327 (vect_analyze_loop): Initialize n_stmts.
1329 2018-06-26 Richard Biener <rguenther@suse.de>
1332 * fold-const.c (fold_convertible_p): Pointer extension
1335 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
1338 * dwarf2out.c (loc_list_has_views): Adjust comments.
1339 (dw_loc_list): Split single cross-partition range with
1342 2018-06-25 Jeff Law <law@redhat.com>
1344 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
1345 on -mbig-switch by default.
1347 * config/v850/predicates.md (const_float_1_operand): Fix match_code
1349 (const_float_0_operand): Remove unused predicate.
1350 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
1351 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
1352 (recipsf2): New expander. Original pattern now called
1354 (recipdf2, recipdf2_insn): Similarly.
1355 (rsqrtsf2, rsqrtsf2_insn): Similarly
1356 (rsqrtdf2, rsqrtdf2_insn): Similarly
1358 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
1360 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
1361 Simplify logic for FreeBSD (twice).
1363 2018-06-25 Martin Sebor <msebor@redhat.com>
1365 PR tree-optimization/86204
1366 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
1367 a strnlen result if it's less than the length of the string.
1369 2018-06-25 Martin Sebor <msebor@redhat.com>
1371 PR tree-optimization/85700
1372 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
1373 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
1374 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
1376 2018-06-25 Martin Sebor <msebor@redhat.com>
1378 * doc/extend.texi (Zero-length arrays): Update and clarify.
1380 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
1382 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
1383 added IEEE/IBM long double multilib support on PowerPC little
1384 endian Linux systems.
1385 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
1386 (MULTILIB_DEFAULTS): Likewise.
1387 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1389 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
1390 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
1391 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
1393 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
1396 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
1397 (REORDER_45): Likewise.
1399 2018-06-25 Jeff Law <law@redhat.com>
1401 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
1402 dividend to 32 bits. Adjust length.
1403 (udivmodhi4): Cleanup output template. Fix length.
1405 2018-06-25 Carl Love <cel@us.ibm.com>
1407 * config/rs6000/vsx.md: Change word selector to prefered location.
1409 2018-06-25 Richard Biener <rguenther@suse.de>
1411 PR tree-optimization/86304
1412 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
1413 epilogue-if-converted loops as well.
1415 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
1417 * lto-section-out.c (lto_begin_section): Do not print section
1418 name for noaddr and unnumbered dumps.
1420 2018-06-25 Richard Biener <rguenther@suse.de>
1422 * tree-vectorizer.h (struct vec_info_shared): New structure
1423 with parts split out from struct vec_info and loop_nest from
1424 struct _loop_vec_info.
1425 (struct vec_info): Adjust accordingly.
1426 (struct _loop_vec_info): Likewise.
1427 (LOOP_VINFO_LOOP_NEST): Adjust.
1428 (LOOP_VINFO_DATAREFS): Likewise.
1429 (LOOP_VINFO_DDRS): Likewise.
1430 (struct _bb_vec_info): Likewise.
1431 (BB_VINFO_DATAREFS): Likewise.
1432 (BB_VINFO_DDRS): Likewise.
1433 (struct _stmt_vec_info): Add dr_aux member.
1434 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
1435 (DR_MISALIGNMENT_UNINITIALIZED): New.
1436 (set_dr_misalignment): Adjust.
1437 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
1438 (vect_analyze_loop): Adjust prototype.
1439 (vect_analyze_loop_form): Likewise.
1440 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
1441 Compute dependences lazily.
1442 (vect_record_base_alignments): Use shared datarefs/ddrs.
1443 (vect_verify_datarefs_alignment): Likewise.
1444 (vect_analyze_data_refs_alignment): Likewise.
1445 (vect_analyze_data_ref_accesses): Likewise.
1446 (vect_analyze_data_refs): Likewise.
1447 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
1448 constructor parameter for shared part.
1449 (vect_analyze_loop_form): Pass in shared part and adjust.
1450 (vect_analyze_loop_2): Pass in storage for the number of
1451 stmts. Move loop nest finding to the caller. Compute
1453 (vect_analyze_loop): Pass in shared part.
1454 (vect_transform_loop): Verify shared datarefs are unchanged.
1455 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
1456 constructor parameter for shared part.
1457 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
1458 (vect_slp_bb): Verify shared datarefs are unchanged before
1460 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
1462 (new_stmt_vec_info): Initialize DR_AUX misalignment to
1463 DR_MISALIGNMENT_UNINITIALIZED.
1464 * tree-vectorizer.c (vec_info::vec_info): Add constructor
1465 parameter for shared part.
1466 (vec_info::~vec_info): Adjust.
1467 (vec_info_shared::vec_info_shared): New.
1468 (vec_info_shared::~vec_info_shared): Likewise.
1469 (vec_info_shared::save_datarefs): Likewise.
1470 (vec_info_shared::check_datarefs): Likewise.
1471 (try_vectorize_loop_1): Construct shared part live for analyses
1472 of a single loop for multiple vector sizes.
1473 * tree-parloops.c (gather_scalar_reductions): Adjust.
1475 2018-06-25 Richard Biener <rguenther@suse.de>
1477 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
1478 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
1479 (vect_analyze_data_refs): Remove similar code from here and
1480 simplify accordingly.
1482 2018-06-25 Richard Biener <rguenther@suse.de>
1484 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
1485 for reverse storage order accesses rather than asserting
1486 they cannot happen here.
1488 2018-06-25 Tom de Vries <tdevries@suse.de>
1491 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
1492 Use data16 instead of .byte for insn prefix.
1494 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
1497 * parser.c (make_char_string_pack): Pass this literal chars
1498 through cpp_interpret_string.
1499 (cp_parser_userdef_numeric_literal): Check the result of
1500 make_char_string_pack.
1502 2018-06-24 Maya Rashish <coypu@sdf.org>
1504 * ginclude/stddef.h: Simplify conditions around avoiding
1505 re-definition of __size_t.
1507 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1509 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
1510 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
1512 2018-06-22 Maya Rashish <coypu@sdf.org>
1514 * doc/invoke.texi (mno-fancy-math-387): Update for changes
1515 made to OpenBSD and NetBSD through the years.
1517 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1519 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1520 behavior of vec_pack (vector double, vector double) to match
1521 behavior of vec_float2 (vector double, vector double).
1523 2018-06-22 Olivier Hainque <hainque@adacore.com>
1525 * gimplify.c (gimplify_function_tree): Prevent creation
1526 of a trampoline for the address of the current function
1527 passed to entry/exit instrumentation hooks.
1529 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
1532 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
1535 2018-06-22 Martin Liska <mliska@suse.cz>
1537 PR tree-optimization/86263
1538 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1539 Bail out if is_enabled is false.
1540 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
1542 (jump_table_cluster::is_enabled): New function.
1544 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1546 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1547 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
1548 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
1549 (lto_input_ts_binfo_tree_pointers): Likewise.
1550 * tree-streamer-out.c (streamer_write_tree_bitfields,
1551 write_ts_binfo_tree_pointers): Likewise.
1552 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
1554 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
1556 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
1558 2018-06-22 Martin Liska <mliska@suse.cz>
1560 * symbol-summary.h (get): Make it pure and inline move
1561 functionality from ::get function.
1562 (get): Remove and inline into ::get and ::get_create.
1563 (get_create): Move code from ::get function.
1565 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1568 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
1569 -x assembler-with-cpp.
1571 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1573 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
1574 _FILE_OFFSET_BITS=64 for C++.
1576 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
1578 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
1579 conversion insn that shows up when pr85657-3.c is compiled using
1580 IEEE 128-bit long double.
1581 (neg<mode>2_internal): Use the correct mode to check whether the
1582 mode is IBM extended.
1583 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
1584 multiply and divide external functions from being created more
1587 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1589 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
1591 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
1592 the edge can be forwarded.
1593 (cfg_layout_merge_blocks): Likewise.
1595 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1597 * except.c (finish_eh_generation): Commit edge insertions only after
1598 the EH edges have been redirected from post-landing to landing pads.
1600 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1602 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
1603 create_tmp_var_for to create the FRAME decl.
1604 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
1606 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
1608 * tree-inline.c (copy_edges_for_bb): Minor tweak.
1609 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
1610 debug statement when resetting its value.
1611 (expand_call_inline): Copy the locus of the call onto the assignment
1612 of the return value, if any. Use local variable in more cases.
1614 2018-06-21 Martin Liska <mliska@suse.cz>
1616 * ipa-pure-const.c (propagate_nothrow): Use
1617 funct_state_summaries->get.
1618 (dump_malloc_lattice): Likewise.
1619 (propagate_malloc): Likewise.
1621 2018-06-21 Richard Biener <rguenther@suse.de>
1623 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
1624 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
1625 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
1627 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
1628 BLOCK_ABSTRACT_ORIGIN unconditionally.
1630 2018-06-21 David Malcolm <dmalcolm@redhat.com>
1632 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
1634 * ipa-reference.c (ipa_reference_c_finalize): Delete
1635 ipa_ref_opt_sum_summaries and set it to NULL.
1637 2018-06-21 Tom de Vries <tdevries@suse.de>
1639 PR tree-optimization/85859
1640 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
1641 test with comment from bb_no_side_effects_p.
1643 2018-06-21 Richard Biener <rguenther@suse.de>
1645 PR tree-optimization/86232
1646 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
1647 max for constant niter.
1649 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1651 * config/aarch64/aarch64-simd.md
1652 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
1654 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
1656 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
1657 Make opernads of the unspec commutative.
1659 2018-06-21 Richard Biener <rguenther@suse.de>
1661 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
1662 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
1663 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
1664 (vect_analyze_data_ref_dependence): Re-order checks to deal with
1666 (vect_record_base_alignments): Do not record base alignment
1667 for gathers or scatters.
1668 (vect_compute_data_ref_alignment): Drop return value that is always
1669 true. Bail out early for gathers or scatters.
1670 (vect_enhance_data_refs_alignment): Bail out early for gathers
1672 (vect_find_same_alignment_drs): Likewise.
1673 (vect_analyze_data_refs_alignment): Remove dead code.
1674 (vect_slp_analyze_and_verify_node_alignment): Likewise.
1675 (vect_analyze_data_refs): For possible gathers or scatters do
1676 not create an alternate DR, just check their possible validity
1677 and mark them. Adjust DECL_NONALIASED handling to not rely
1679 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
1680 update inits of gathers or scatters.
1681 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
1682 Also copy gather/scatter flag to pattern vinfo.
1684 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
1686 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
1687 behavior of vec_packsu (vector unsigned long long, vector unsigned
1688 long long) to match behavior of vec_packs with same signature.
1690 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
1691 Thomas Schwinge <thomas@codesourcery.com>
1692 Cesar Philippidis <cesar@codesourcery.com>
1694 * gimplify.c (gimplify_scan_omp_clauses): Add support for
1695 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1696 (gimplify_adjust_omp_clauses): Likewise.
1697 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
1698 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
1699 (gimplify_omp_target_update): Update handling of acc update and
1701 * omp-low.c (install_var_field): Remove unused parameter
1702 base_pointers_restrict.
1703 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
1704 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
1706 (omp_target_base_pointers_restrict_p): Delete.
1707 (scan_omp_target): Update call to scan_sharing_clauses.
1708 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
1710 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1711 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
1712 (convert_local_omp_clauses): Likewise.
1713 * tree-pretty-print.c (dump_omp_clause): Likewise.
1714 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
1716 (omp_clause_code_name): Likewise.
1718 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1721 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
1724 PR tree-optimization/86231
1725 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
1726 anti-range don't overwrite *vr0min before using it to compute *vr0max.
1728 2018-06-20 Tom de Vries <tdevries@suse.de>
1730 PR tree-optimization/86097
1731 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
1732 iv type if signedness of iv type is not the same as that of *nit.
1734 2018-06-20 Jakub Jelinek <jakub@redhat.com>
1736 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
1737 EDGE_EH edges, verify they are all EDGE_EH.
1739 2018-06-20 Maya Rashish <coypu@sdf.org>
1741 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
1743 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1745 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
1746 * config/aarch64/aarch64.c (xgene1_tunings): Add
1747 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
1748 (aarch64_mode_valid_for_sched_fusion_p):
1749 Allow 16-byte modes.
1750 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
1751 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
1753 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
1755 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
1756 * config/aarch64/iterators.md (VQ2): New mode iterator.
1758 2018-06-20 Martin Liska <mliska@suse.cz>
1760 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1761 Change default ratio from 10 to 8.
1763 2018-06-20 Martin Liska <mliska@suse.cz>
1765 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1767 (bit_test_cluster::find_bit_tests): Likewise.
1768 (switch_decision_tree::analyze_switch_statement): Find clusters.
1769 * tree-switch-conversion.h (struct jump_table_cluster): Document
1772 2018-06-20 Martin Liska <mliska@suse.cz>
1774 * tree-switch-conversion.c (switch_conversion::collect):
1775 Record m_uniq property.
1776 (switch_conversion::expand): Bail out for special conditions.
1777 (group_cluster::~group_cluster): New.
1778 (group_cluster::group_cluster): Likewise.
1779 (group_cluster::dump): Likewise.
1780 (jump_table_cluster::emit): New.
1781 (switch_decision_tree::fix_phi_operands_for_edges): New.
1782 (struct case_node): Remove struct.
1783 (jump_table_cluster::can_be_handled): New.
1784 (case_values_threshold): Moved to header.
1785 (reset_out_edges_aux): Likewise.
1786 (jump_table_cluster::is_beneficial): New.
1787 (bit_test_cluster::can_be_handled): Likewise.
1788 (add_case_node): Remove.
1789 (bit_test_cluster::is_beneficial): New.
1790 (case_bit_test::cmp): New.
1791 (bit_test_cluster::emit): New.
1792 (expand_switch_as_decision_tree_p): Remove.
1793 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
1794 (fix_phi_operands_for_edge): Likewise.
1795 (switch_decision_tree::analyze_switch_statement): New.
1796 (compute_cases_per_edge): Move ...
1797 (switch_decision_tree::compute_cases_per_edge): ... here.
1798 (try_switch_expansion): Likewise.
1799 (switch_decision_tree::try_switch_expansion): Likewise.
1800 (record_phi_operand_mapping): Likewise.
1801 (switch_decision_tree::record_phi_operand_mapping): Likewise.
1802 (emit_case_decision_tree): Likewise.
1803 (switch_decision_tree::emit): Likewise.
1804 (balance_case_nodes): Likewise.
1805 (switch_decision_tree::balance_case_nodes): Likewise.
1806 (dump_case_nodes): Likewise.
1807 (switch_decision_tree::dump_case_nodes): Likewise.
1808 (emit_jump): Likewise.
1809 (switch_decision_tree::emit_jump): Likewise.
1810 (emit_cmp_and_jump_insns): Likewise.
1811 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1812 (emit_case_nodes): Likewise.
1813 (switch_decision_tree::emit_case_nodes): Likewise.
1814 (conditional_probability): Remove.
1815 * tree-switch-conversion.h (enum cluster_type): New.
1817 (struct cluster): Likewise.
1818 (cluster::cluster): Likewise.
1819 (struct simple_cluster): Likewise.
1820 (simple_cluster::simple_cluster): Likewise.
1821 (struct group_cluster): Likewise.
1822 (struct jump_table_cluster): Likewise.
1823 (struct bit_test_cluster): Likewise.
1824 (struct min_cluster_item): Likewise.
1825 (struct case_tree_node): Likewise.
1826 (case_tree_node::case_tree_node): Likewise.
1827 (jump_table_cluster::case_values_threshold): Likewise.
1828 (struct case_bit_test): Likewise.
1829 (struct switch_decision_tree): Likewise.
1830 (struct switch_conversion): Likewise.
1831 (switch_decision_tree::reset_out_edges_aux): Likewise.
1833 2018-06-20 Martin Liska <mliska@suse.cz>
1835 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
1836 (hoist_edge_and_branch_if_true): Likewise.
1837 (expand_switch_using_bit_tests_p): Likewise.
1838 (struct case_bit_test): Likewise.
1839 (case_bit_test_cmp): Likewise.
1840 (emit_case_bit_tests): Likewise.
1841 (switch_conversion::switch_conversion): New class.
1842 (struct switch_conv_info): Remove old struct.
1843 (collect_switch_conv_info): More to ...
1844 (switch_conversion::collect): ... this.
1845 (check_range): Likewise.
1846 (switch_conversion::check_range): Likewise.
1847 (check_all_empty_except_final): Likewise.
1848 (switch_conversion::check_all_empty_except_final): Likewise.
1849 (check_final_bb): Likewise.
1850 (switch_conversion::check_final_bb): Likewise.
1851 (create_temp_arrays): Likewise.
1852 (switch_conversion::create_temp_arrays): Likewise.
1853 (free_temp_arrays): Likewise.
1854 (gather_default_values): Likewise.
1855 (switch_conversion::gather_default_values): Likewise.
1856 (build_constructors): Likewise.
1857 (switch_conversion::build_constructors): Likewise.
1858 (constructor_contains_same_values_p): Likewise.
1859 (switch_conversion::contains_same_values_p): Likewise.
1860 (array_value_type): Likewise.
1861 (switch_conversion::array_value_type): Likewise.
1862 (build_one_array): Likewise.
1863 (switch_conversion::build_one_array): Likewise.
1864 (build_arrays): Likewise.
1865 (switch_conversion::build_arrays): Likewise.
1866 (gen_def_assigns): Likewise.
1867 (switch_conversion::gen_def_assigns): Likewise.
1868 (prune_bbs): Likewise.
1869 (switch_conversion::prune_bbs): Likewise.
1870 (fix_phi_nodes): Likewise.
1871 (switch_conversion::fix_phi_nodes): Likewise.
1872 (gen_inbound_check): Likewise.
1873 (switch_conversion::gen_inbound_check): Likewise.
1874 (process_switch): Use the newly created class.
1875 (switch_conversion::expand): New.
1876 (switch_conversion::~switch_conversion): New.
1877 * tree-switch-conversion.h: New file.
1879 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1881 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
1882 tree-vect-patterns.c.
1883 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
1884 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
1885 (vect_recog_sad_pattern): Likewise.
1886 (vect_recog_widen_sum_pattern): Likewise.
1887 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
1888 (vect_recog_widen_shift_pattern): Remove the type_in argument.
1889 (vect_recog_rotate_pattern): Likewise.
1890 (vect_recog_mult_pattern): Likewise.
1891 (vect_recog_vector_vector_shift_pattern): Likewise.
1892 (vect_recog_divmod_pattern): Likewise.
1893 (vect_recog_mixed_size_cond_pattern): Likewise.
1894 (vect_recog_bool_pattern): Likewise.
1895 (vect_recog_mask_conversion_pattern): Likewise.
1896 (vect_try_gather_scatter_pattern): Likewise.
1897 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
1898 (vect_recog_over_widening_pattern): Likewise.
1899 (vect_recog_gather_scatter_pattern): Likewise.
1900 (vect_recog_func_ptr): Move from tree-vectorizer.h
1901 (vect_vect_recog_func_ptrs): Move further down the file.
1902 (vect_recog_func): Likewise. Remove the third argument.
1903 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
1904 (vect_pattern_recog_1): Expect the pattern function to do any
1905 necessary target tests. Also expect it to provide a vector type.
1906 Remove the type_in handling.
1908 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1910 * tree-vect-patterns.c (vect_pattern_detected): New function.
1911 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
1912 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
1913 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
1914 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
1915 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1916 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1917 (vect_recog_mask_conversion_pattern)
1918 (vect_try_gather_scatter_pattern): Likewise.
1920 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1922 * tree-vect-patterns.c (vect_get_internal_def): New function.
1923 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1924 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
1925 (search_type_for_mask_1): Use it.
1927 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1929 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1930 redundant WIDEN_SUM_EXPR handling.
1931 (vect_recog_sad_pattern): Likewise.
1933 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1935 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
1936 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
1937 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
1938 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
1939 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
1941 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1943 * tree-vect-stmts.c (vectorizable_call): Make sure that we
1944 use the stmt_vec_info of the original bb statement for the
1945 new zero assignment, even if the call is part of a pattern.
1947 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1949 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
1950 that the sequence is attached to the original statement rather
1951 than the pattern statement.
1952 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
1953 PATTERN_DEF_SEQ from the original statement rather than
1954 the main pattern statement.
1955 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
1956 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
1957 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
1959 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
1961 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
1962 definition statements before the early exit for statements that aren't
1964 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
1966 (vect_transform_loop): ...here. Process pattern definition
1967 statements without first checking whether the main pattern
1968 statement is live or relevant.
1970 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
1972 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
1973 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
1975 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
1977 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
1978 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
1979 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
1980 (expand_block_compare): Change select_block_compare_mode call.
1981 (expand_strncmp_align_check): Use new functions, fix comment.
1982 (emit_final_str_compare_gpr): New function.
1983 (expand_strn_compare): Refactor and clean up code.
1984 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
1986 2018-06-19 Tony Reix <tony.reix@atos.com>
1987 Damien Bergamini <damien.bergamini@atos.com>
1988 David Edelsohn <dje.gcc@gmail.com>
1990 * collect2.c (static_obj): New variable.
1991 (static_libs): New variable.
1992 (is_in_list): Uncomment declaration.
1993 (main): Track AIX libraries linked statically.
1994 (is_in_list): Uncomment definition.
1995 (scan_prog_file): Don't add AIX shared libraries initializer
1996 to constructor list if linking statically.
1998 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2000 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2002 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2004 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2006 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2009 2018-06-19 Martin Liska <mliska@suse.cz>
2011 * config/i386/i386.c (ix86_can_inline_p): Do not use
2012 ipa_fn_summaries::get_create.
2013 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2015 (devirtualization_time_bonus): Likewise.
2016 (ipcp_propagate_stage): Likewise.
2017 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2018 (edge_set_predicate): Likewise.
2019 (evaluate_conditions_for_known_args): Likewise.
2020 (evaluate_properties_for_edge): Likewise.
2021 (ipa_call_summary::reset): Tranform to ...
2022 (ipa_call_summary::~ipa_call_summary): ... this.
2023 (ipa_fn_summary::reset): Transform to ...
2024 (ipa_fn_summary::~ipa_fn_summary): ... this.
2025 (ipa_fn_summary_t::remove): Rename to ...
2026 (ipa_fn_summary_t::remove_callees): ... this.
2027 (ipa_fn_summary_t::duplicate): Use placement new
2028 instead of memory copy.
2029 (ipa_call_summary_t::duplicate): Likewise.
2030 (ipa_call_summary_t::remove): Remove.
2031 (dump_ipa_call_summary): Change get_create to get.
2032 (ipa_dump_fn_summary): Dump only when summary exists.
2033 (analyze_function_body): Use symbol_summary::get instead
2035 (compute_fn_summary): Likewise.
2036 (estimate_edge_devirt_benefit): Likewise.
2037 (estimate_edge_size_and_time): Likewise.
2038 (inline_update_callee_summaries): Likewise.
2039 (remap_edge_change_prob): Likewise.
2040 (remap_edge_summaries): Likewise.
2041 (ipa_merge_fn_summary_after_inlining): Likewise.
2042 (write_ipa_call_summary): Likewise.
2043 (ipa_fn_summary_write): Likewise.
2044 (ipa_free_fn_summary): Likewise.
2045 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2046 (struct ipa_call_summary): Likewise.
2047 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2049 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2050 (estimate_size_after_inlining): Likewise.
2051 (estimate_growth): Likewise.
2052 (growth_likely_positive): Likewise.
2053 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2054 (inline_call): Likewise.
2055 * ipa-inline.c (caller_growth_limits): Likewise.
2056 (can_inline_edge_p): Likewise.
2057 (can_inline_edge_by_limits_p): Likewise.
2058 (compute_uninlined_call_time): Likewise.
2059 (compute_inlined_call_time): Likewise.
2060 (want_inline_small_function_p): Likewise.
2061 (edge_badness): Likewise.
2062 (update_caller_keys): Likewise.
2063 (update_callee_keys): Likewise.
2064 (inline_small_functions): Likewise.
2065 (inline_to_all_callers_1): Likewise.
2066 (dump_overall_stats): Likewise.
2067 (early_inline_small_functions): Likewise.
2068 (early_inliner): Likewise.
2069 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2070 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2071 * ipa-pure-const.c (malloc_candidate_p): Likewise.
2072 * ipa-split.c (execute_split_functions): Likewise.
2073 * symbol-summary.h: Likewise.
2074 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
2076 2018-06-19 Richard Biener <rguenther@suse.de>
2078 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
2079 (vectorize_loops): ... here. Fix dbgcnt handling.
2080 (try_vectorize_loop): Wrap try_vectorize_loop_1.
2082 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
2085 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
2086 ieee128 argument takes up only one (vector) register, not two (floating
2089 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2091 * gimplify.c (gimplify_init_constructor): Really never clear for an
2092 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
2094 2018-06-19 Richard Biener <rguenther@suse.de>
2096 PR tree-optimization/86179
2097 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
2098 after failed recognition.
2100 2018-06-18 Martin Sebor <msebor@redhat.com>
2103 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2104 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2105 Handle integer subtraction.
2106 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2107 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2109 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2111 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
2112 param from rtx to rtx_insn *.
2113 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
2115 (frv_ifcvt_modify_insn): Likwise.
2116 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
2117 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
2118 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
2119 as_a <rtx_insn *> cast to local "unprotected_region" once
2120 it's been established that it's not NULL or pc_rtx.
2121 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
2122 param "sethi" from rtx to rtx_insn *.
2123 (nds32_group_float_insns): Likewise for param "insn".
2124 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
2126 (vax_output_int_subtract): Likewise.
2127 * config/vax/vax.c (vax_output_int_add): Likewise for param
2129 (vax_output_int_subtract): Likewise.
2130 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
2131 (emit_pattern_after): Likewise for param "after".
2132 (emit_insn_after): Likewise.
2133 (emit_jump_insn_after): Likewise.
2134 (emit_call_insn_after): Likewise.
2135 (emit_debug_insn_after): Likewise.
2136 (emit_pattern_before): Likewise for param "before".
2137 (emit_insn_before): Likewise.
2138 (emit_jump_insn_before): Likewise.
2139 * final.c (get_insn_template): Likewise for param "insn", removing
2141 * output.h (get_insn_template): Likewise for 2nd param.
2142 * rtl.h (emit_insn_before): Likewise.
2143 (emit_jump_insn_before): Likewise.
2144 (emit_debug_insn_before_noloc): Likewise.
2145 (emit_insn_after): Likewise.
2146 (emit_jump_insn_after): Likewise.
2147 (emit_call_insn_after): Likewise.
2148 (emit_debug_insn_after): Likewise.
2149 (set_insn_deleted): Likewise for param.
2151 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
2154 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
2155 floating point modes, so that IFmode is numerically greater than
2156 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
2157 to declare the ordering. This prevents IFmode from being
2158 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
2159 machine. Include rs6000-modes.h to share the fractional values
2160 between genmodes* and the rest of the compiler.
2164 * config/rs6000/rs6000-modes.h: New file.
2165 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
2166 meaning of rs6000_long_double_size so that 126..128 selects an
2167 appropriate 128-bit floating point type.
2168 (rs6000_option_override_internal): Likewise.
2169 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
2170 (TARGET_LONG_DOUBLE_128): Change the meaning of
2171 rs6000_long_double_size so that 126..128 selects an appropriate
2172 128-bit floating point type.
2173 (LONG_DOUBLE_TYPE_SIZE): Update comment.
2174 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
2175 source and destination to match the standard usage.
2176 (truncifkf2): Likewise.
2177 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
2178 ISA 2.07 to use an explicit clobber, instead of passing in a
2180 (copysign<mode>3_soft): Likewise.
2182 2018-06-18 David Malcolm <dmalcolm@redhat.com>
2184 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2185 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
2186 (vect_slp_analyze_instance_dependence): Likewise.
2187 (vect_enhance_data_refs_alignment): Likewise.
2188 (vect_analyze_data_refs_alignment): Likewise.
2189 (vect_slp_analyze_and_verify_instance_alignment
2190 (vect_analyze_data_ref_accesses): Likewise.
2191 (vect_prune_runtime_alias_test_list): Likewise.
2192 (vect_analyze_data_refs): Likewise.
2193 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
2194 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
2195 (vect_analyze_scalar_cycles_1): Likewise.
2196 (vect_get_loop_niters): Likewise.
2197 (vect_analyze_loop_form_1): Likewise.
2198 (vect_update_vf_for_slp): Likewise.
2199 (vect_analyze_loop_operations): Likewise.
2200 (vect_analyze_loop): Likewise.
2201 (vectorizable_induction): Likewise.
2202 (vect_transform_loop): Likewise.
2203 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
2204 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2205 (vect_make_slp_decision): Likewise.
2206 (vect_detect_hybrid_slp): Likewise.
2207 (vect_slp_analyze_operations): Likewise.
2208 (vect_slp_bb): Likewise.
2209 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
2210 (vectorizable_bswap): Likewise.
2211 (vectorizable_call): Likewise.
2212 (vectorizable_simd_clone_call): Likewise.
2213 (vectorizable_conversion): Likewise.
2214 (vectorizable_assignment): Likewise.
2215 (vectorizable_shift): Likewise.
2216 (vectorizable_operation): Likewise.
2217 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
2219 2018-06-18 Martin Sebor <msebor@redhat.com>
2221 PR tree-optimization/81384
2222 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
2223 * builtins.c (expand_builtin_strnlen): New function.
2224 (expand_builtin): Call it.
2225 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
2226 * builtins.def (BUILT_IN_STRNLEN): New.
2227 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
2228 Warn for bounds in excess of maximum object size.
2229 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
2230 single-value ranges. Handle strnlen.
2231 (handle_builtin_strlen): Handle strnlen.
2232 (strlen_check_and_optimize_stmt): Same.
2233 * doc/extend.texi (Other Builtins): Document strnlen.
2235 2018-06-18 Maya Rashish <coypu@sdf.org>
2237 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
2238 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
2239 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
2241 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
2243 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
2245 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2247 * tree.c (escaped_string::escape): Replace cast to char * by
2248 const_cast<char *> (unescaped).
2250 2018-06-18 Nick Clifton <nickc@redhat.com>
2253 * tree.c (escaped_string): New class. Converts an unescaped
2254 string into its escaped equivalent.
2255 (warn_deprecated_use): Use the new class to convert the
2256 deprecation message, if present.
2257 (test_escaped_strings): New self test.
2258 (test_c_tests): Add test_escaped_strings.
2259 * doc/extend.texi (deprecated): Add a note that the
2260 deprecation message is affected by the -fmessage-length
2261 option, and that control characters will be escaped.
2262 (#pragma GCC error): Document this pragma.
2263 (#pragma GCC warning): Likewise.
2264 * doc/invoke.texi (-fmessage-length): Document this option's
2265 effect on the #warning and #error preprocessor directives and
2266 the deprecated attribute.
2268 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
2270 * tree.c (decl_value_expr_lookup): Revert latest change.
2271 (decl_value_expr_insert): Likewise.
2273 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
2275 * gimplify.c (nonlocal_vlas): Delete.
2276 (nonlocal_vla_vars): Likewise.
2277 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
2279 (gimplify_body): Do not create and destroy nonlocal_vlas.
2280 * tree-nested.c: Include diagnostic.h.
2281 (use_pointer_in_frame): Tweak.
2282 (lookup_field_for_decl): Add assertion and declare the transformation.
2283 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
2284 internal error when the reference is in a wrong context. Do not
2285 create a debug decl by default.
2286 (note_nonlocal_block_vlas): Delete.
2287 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
2288 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
2289 create a debug decl by default.
2290 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
2291 call is in a wrong context.
2292 (fixup_vla_decls): New function.
2293 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
2294 debug variables were created.
2295 * tree.c (decl_value_expr_lookup): Add checking assertion.
2296 (decl_value_expr_insert): Likewise.
2298 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2301 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
2302 * tree-scalar-evolution.c (interpret_expr): Likewise.
2303 (expression_expensive_p): Likewise.
2304 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
2305 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
2306 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
2307 (ssa_defined_by_minus_one_stmt_p): New.
2309 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
2312 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
2313 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
2314 * dojump.c (do_jump): Likewise.
2315 * expr.c (expand_expr_real_2): Check operand type's sign.
2316 * fold-const.c (const_unop): Handle ABSU_EXPR.
2317 (fold_abs_const): Likewise.
2318 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
2319 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
2320 (strip_sign_op_1): Likesise.
2321 * match.pd: Add new pattern to generate ABSU_EXPR.
2322 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
2323 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2324 * tree-eh.c (operation_could_trap_helper_p): Likewise.
2325 * tree-inline.c (estimate_operator_cost): Likewise.
2326 * tree-pretty-print.c (dump_generic_node): Likewise.
2327 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
2328 * tree.def (ABSU_EXPR): New.
2330 2018-06-16 Jakub Jelinek <jakub@redhat.com>
2333 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
2334 documented as preserved for backward compatibility only.
2335 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
2337 PR rtl-optimization/86108
2338 * bb-reorder.c (create_forwarder_block): Renamed to ...
2339 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
2340 jump from new landing pad to the second part.
2341 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
2344 2018-06-15 Jakub Jelinek <jakub@redhat.com>
2347 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
2348 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
2349 Only call store_expr for halves if the mode is the same.
2352 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
2353 Fix up comment formatting.
2355 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
2357 * typed-splay-tree.h (typed_splay_tree::remove): New function.
2358 (typed_splay_tree::closure,
2359 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
2360 (typed_splay_tree::typed_splay_tree,
2361 typed_splay_tree::operator =): Declared private.
2362 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
2363 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
2364 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
2365 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
2366 typed_splay_tree::splay_tree_splay,
2367 typed_splay_tree::splay_tree_foreach_helper,
2368 typed_splay_tree::splay_tree_insert,
2369 typed_splay_tree::splay_tree_remove,
2370 typed_splay_tree::splay_tree_lookup,
2371 typed_splay_tree::splay_tree_predecessor,
2372 typed_splay_tree::splay_tree_successor,
2373 typed_splay_tree::splay_tree_min,
2374 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
2375 (typed_splay_tree::root, typed_splay_tree::comp,
2376 typed_splay_tree::delete_key,
2377 typed_splay_tree::delete_value): New data members.
2378 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
2379 typed_splay_tree::remove.
2381 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
2383 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
2384 -mginv and -mno-ginv to the assembler.
2385 * config/mips/mips.opt (-mcrc): New option.
2387 * doc/invoke.text (-mcrc): Document.
2390 2018-06-15 Nick Clifton <nickc@redhat.com>
2393 * tree.c (escaped_string): New class. Converts an unescaped
2394 string into its escaped equivalent.
2395 (warn_deprecated_use): Use the new class to convert the
2396 deprecation message, if present.
2397 (test_escaped_strings): New self test.
2398 (test_c_tests): Add test_escaped_strings.
2399 * doc/extend.texi (deprecated): Add a note that the
2400 deprecation message is affected by the -fmessage-length
2401 option, and that control characters will be escaped.
2402 (#pragma GCC error): Document this pragma.
2403 (#pragma GCC warning): Likewise.
2404 * doc/invoke.texi (-fmessage-length): Document this option's
2405 effect on the #warning and #error preprocessor directives and
2406 the deprecated attribute.
2408 2018-06-15 Richard Biener <rguenther@suse.de>
2410 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
2411 here, also noting vector size used.
2412 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
2413 size used in MSG_OPTIMIZED_LOCATIONS dump.
2414 (pass_slp_vectorize::execute): Adjust.
2416 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
2419 * config/arc/arc.c (arc_return_address_register): Fix
2422 2018-06-15 Richard Biener <rguenther@suse.de>
2425 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
2426 leave useless conversion stripping to force_gimple_operand_gsi.
2427 (gimplify_build2): Likewise.
2428 (gimplify_build1): Likewise.
2430 2018-06-15 Richard Biener <rguenther@suse.de>
2433 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2434 before adjusting their block.
2436 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2438 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
2439 multilibs for *-*-rtems*.
2440 * config/riscv/t-rtems: New file.
2442 2018-06-14 Jakub Jelinek <jakub@redhat.com>
2445 * match.pd ((A +- CST1) +- CST2): Punt if last resort
2446 unsigned_type_for returns NULL.
2449 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
2450 subregs of multi-word pseudos unless the float mode has word size.
2452 2018-06-14 Richard Biener <rguenther@suse.de>
2455 * tree-vect-generic.c (build_word_mode_vector_type): Remove
2456 duplicate and harmful type_hash_canon.
2457 * tree.c (type_hash_canon): Assert we didn't find ourselves.
2459 2018-06-14 Richard Biener <rguenther@suse.de>
2462 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
2465 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
2467 * config/rtems.h (STDINT_LONG32): Define.
2469 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
2470 Prachi Godbole <prachi.godbole@imgtec.com>
2472 * config/mips/mips-cpus.def: Define P6600.
2473 * config/mips/mips-tables.opt: Regenerate.
2474 * config/mips/mips.c (mips_ucbranch_type): New enum.
2475 (mips_rtx_cost_data): Add support for P6600.
2476 (mips_issue_rate): Likewise.
2477 (mips_multipass_dfa_lookahead): Likewise.
2478 (mips_avoid_hazard): Likewise.
2479 (mips_reorg_process_insns): Likewise.
2480 (mips_classify_branch_p6600): New function.
2481 * config/mips/mips.h (TUNE_P6600): New define.
2482 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
2483 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
2484 * config/mips/mips.md: Include p6600.md.
2485 (processor): Add p6600.
2486 * config/mips/p6600.md: New file.
2487 * doc/invoke.texi: Add p6600 to supported architectures.
2489 2018-06-13 Martin Sebor <msebor@redhat.com>
2491 PR tree-optimization/86114
2492 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
2494 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
2496 2018-06-13 Richard Biener <rguenther@suse.de>
2498 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
2499 Properly set vector type of the intermediate stmt.
2500 * tree-vect-stmts.c (vectorizable_operation): The destination
2501 var always has vectype_out type.
2503 2018-06-13 Jeff Law <law@redhat.com>
2505 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
2506 integer 0 for argument to print_rtl_with_bb.
2507 (rl78_reorg): Likewise.
2509 2018-06-13 David Malcolm <dmalcolm@redhat.com>
2511 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
2512 from rtx to rtx_insn *.
2513 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
2515 (add_sched_insns_for_speculation): Likewise for local "target",
2516 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2517 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
2518 from rtx_insn ** to rtx_code_label **.
2519 (reorg_emit_nops): Likewise.
2520 (c6x_reorg): Likewise for local "call_labels".
2521 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
2523 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
2524 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
2525 the loops over LABEL_REFS.
2526 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
2528 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
2529 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
2530 (split_branches): Strengthen local "olabel" from rtx to
2531 rtx_insn *, adding a safe_as_a cast.
2532 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
2534 (add_insn_after): Likewise for first two params.
2535 (add_insn_before): Likewise.
2536 (remove_insn): Likewise for param.
2537 (emit_pattern_before_noloc): Likewise for second and third params.
2538 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
2539 (emit_call_insn_before_noloc): Likewise.
2540 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
2542 (emit_barrier_before): Likewise.
2543 (emit_label_before): Strengthen "label" param from "rtx" to
2544 "rtx_code_label *". Strengthen "before" param from "rtx" to
2546 (emit_insn_after_1): Strengthen "after" param from "rtx" to
2548 (emit_pattern_after_noloc): Likewise.
2549 (emit_insn_after_noloc): Likewise.
2550 (emit_jump_insn_after_noloc): Likewise.
2551 (emit_call_insn_after_noloc): Likewise.
2552 (emit_debug_insn_after_noloc): Likewise.
2553 (emit_barrier_after): Likewise.
2554 (emit_label_after): Likewise for both params.
2555 (emit_pattern_after_setloc): Likewise for "after" param. Convert
2556 "loc" param from "int" to "location_t".
2557 (emit_insn_after_setloc): Likewise.
2558 (emit_jump_insn_after_setloc): Likewise.
2559 (emit_call_insn_after_setloc): Likewise.
2560 (emit_debug_insn_after_setloc): Likewise.
2561 (emit_pattern_before_setloc): Likewise for "before" param. Convert
2562 "loc" param from "int" to "location_t".
2563 (emit_pattern_before): Convert NULL_RTX to NULL.
2564 (emit_insn_before_setloc): Convert "loc" param from "int" to
2566 (emit_jump_insn_before_setloc): Likewise.
2567 (emit_call_insn_before_setloc): Likewise.
2568 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
2569 rtx_insn *. Convert "loc" param from "int" to "location_t".
2570 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
2571 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
2572 Convert 3rd param from "int" to "location_t".
2573 (emit_barrier_before, emit_barrier_after, next_real_insn):
2574 Strengthen param from rtx to rtx_insn *.
2575 (emit_label_before): Strengthen 1st param from "rtx" to
2576 "rtx_code_label *". Strengthen 2nd param from "rtx" to
2578 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
2579 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
2580 Strengthen 2nd param from "rtx" to "rtx_insn *".
2581 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
2582 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
2583 Likewise. Convert 3rd param from "int" to "location_t".
2584 (emit_label_after): Strengthen 1st param from "rtx" to
2586 (next_real_insn, remove_insn): Strengthen param from "rtx" to
2588 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
2589 from "rtx" to "rtx_insn *".
2591 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
2593 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
2594 bodies streamed in with -Q.
2595 * dumpfile.c (dump_files): Add lto-stream-out dump file.
2596 * dumpfile.h (tree_dump_index): Add lto_stream_out.
2597 * gimple-streamer-out.c: Include gimple-pretty-print.h
2598 (output_bb): Dump stmts streamed.
2599 * lto-section-out.c: Include print-tree.h
2600 (lto_begin_section): Dump sections created.
2601 (lto_output_decl_index): Dump decl encoded.
2602 * lto-streamer-out.c: Include print-tree.h
2603 (create_output_block): Dump output block created.
2604 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
2605 (output_function): Dump function output.
2606 (output_constructor): Dump constructor streamed.
2607 (write_global_stream): Output indexes encoded.
2608 (produce_asm_for_decls): Dump streams encoded.
2609 * lto-streamer.c (streamer_dump_file): New global var.
2610 * lto-streamer.h (streamer_dump_file): Declare.
2611 * passes.c (ipa_write_summaries): Initialize streamer dump.
2612 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
2615 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
2618 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
2619 offsets for register save directives. Emit a second batch of save
2620 directives, if need be, when the function accesses prior frames.
2622 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2624 * config/arc/fpu.md (fmasf4): Force operand to register.
2625 (fnmasf4): Likewise.
2627 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2629 * config/arc/arc-protos.h (arc_pad_return): Remove.
2630 * config/arc/arc.c (machine_function): Remove force_short_suffix
2632 (arc_print_operand): Adjust printing of '&'.
2633 (arc_verify_short): Remove conditional printing of short suffix.
2634 (arc_final_prescan_insn): Remove reference to size_reason.
2635 (pad_return): New function.
2636 (arc_reorg): Call pad_return.
2637 (arc_pad_return): Remove.
2638 (arc_init_machine_status): Remove reference to force_short_suffix.
2639 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2640 (attr length): When attribute iscompact is true force to 2
2641 regardless; in the case of maybe check if we want to force the
2642 instruction to have 4 bytes length.
2643 (nopv): Change it to generate 4 byte long nop as well.
2644 (blockage): New pattern.
2645 (simple_return): Remove call to arc_pad_return.
2646 (p_return_i): Likewise.
2648 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2650 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2652 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2654 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
2657 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2659 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
2660 for ARC700 and ARCv2.
2662 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
2665 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
2666 operands[2] instead of operands[1].
2669 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2671 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
2672 case, check whether the outer register overlaps an unallocatable
2673 register, not just whether it fits the required class.
2675 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
2677 * poly-int.h (can_div_trunc_p): Add new overload in which all values
2679 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
2680 (memrefs_conflict_p): Likewise.
2681 (init_alias_analysis): Likewise.
2682 * cfgexpand.c (expand_debug_expr): Likewise.
2683 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
2684 * cse.c (fold_rtx): Likewise.
2685 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
2686 * expr.c (emit_block_move_hints): Likewise.
2687 (clear_storage_hints, push_block, emit_push_insn): Likewise.
2688 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
2689 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
2690 (emit_group_store): Likewise.
2691 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
2692 to read the PRE/POST_MODIFY increment.
2693 * calls.c (store_one_arg): Use strip_offset.
2694 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
2696 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
2698 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
2699 (simplify_binary_operation_1): Extend CONST_INT handling to
2701 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
2702 than a HOST_WIDE_INT.
2703 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
2705 (adjust_mems, add_stores): Update accodingly.
2706 (vt_canonicalize_addr): Track polynomial offsets.
2707 (emit_note_insn_var_location): Likewise.
2708 (vt_add_function_parameter): Likewise.
2709 (vt_initialize): Likewise.
2711 2018-06-12 Jeff Law <law@redhat.com>
2713 * config.gcc (alpha*-*-freebsd*): Remove.
2714 * config/alpha/freebsd.h: Remove.
2716 2018-06-12 David Malcolm <dmalcolm@redhat.com>
2719 * spellcheck-tree.c (levenshtein_distance): Rename to...
2720 (get_edit_distance): ...this, and update for underlying renaming.
2721 * spellcheck-tree.h (levenshtein_distance): Rename to...
2722 (get_edit_distance): ...this.
2723 * spellcheck.c (levenshtein_distance): Rename to...
2724 (get_edit_distance): ...this. Convert from Levenshtein distance
2725 to Damerau-Levenshtein distance by supporting transpositions of
2726 adjacent characters. Rename "v1" to "v_next" and "v0" to
2728 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
2729 (selftest::test_edit_distance_unit_test_oneway): ...this, and
2730 update for underlying renaming.
2731 (selftest::levenshtein_distance_unit_test): Rename to...
2732 (selftest::test_get_edit_distance_unit): ...this, and update for
2733 underlying renaming.
2734 (selftest::test_find_closest_string): Add example from PR 69968
2735 where transposition helps
2736 (selftest::test_metric_conditions): Update for renaming.
2737 (selftest::test_metric_conditions): Likewise.
2738 (selftest::spellcheck_c_tests): Likewise.
2739 * spellcheck.h (levenshtein_distance): Rename both overloads to...
2740 (get_edit_distance): ...this.
2741 (best_match::consider): Update for renaming.
2743 2018-06-12 Martin Sebor <msebor@redhat.com>
2745 PR tree-optimization/85259
2746 * builtins.c (compute_objsize): Handle constant offsets.
2747 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
2748 true iff a warning has been issued.
2749 * gimple.h (gimple_nonartificial_location): New function.
2750 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
2751 gimple_nonartificial_location and handle -Wno-system-headers.
2752 (handle_builtin_stxncpy): Same.
2754 2018-06-12 Martin Sebor <msebor@redhat.com>
2757 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
2759 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
2761 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
2762 BUILTIN_VEC_XST entries for pointer to double and long long.
2764 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
2767 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2769 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
2772 2018-06-12 Martin Liska <mliska@suse.cz>
2774 * doc/options.texi: Document IntegerRange.
2776 2018-06-12 Martin Liska <mliska@suse.cz>
2778 * config/i386/i386.opt: Make MPX-related options as Deprecated.
2779 * opt-functions.awk: Handle Deprecated flag.
2780 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
2782 (read_cmdline_option): Report warning for a deprecated option.
2783 * opts.h (struct cl_option): Add new field cl_deprecated.
2784 (CL_ERR_DEPRECATED): New.
2786 2018-06-12 Martin Liska <mliska@suse.cz>
2788 * doc/options.texi: Document Deprecated option flag.
2790 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2792 * config/arc/arc-arch.h (arc_extras): New enum.
2793 (arc_cpu_t):Add field extra.
2794 (arc_cpu_types): Consider the extras.
2795 * config/arc/arc-cpus.def: Add extras info.
2796 * config/arc/arc-opts.h (processor_type): Consider extra field.
2797 * config/arc/arc.c (arc_override_options): Handle extra field.
2799 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
2801 * config/arc/arc-arch.h: Update ARC_OPTX macro.
2802 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
2804 * config/arc/arc.c (arc_init): Update pic warning.
2805 (irq_range): Update irq range parsing warnings.
2806 (arc_override_options): Update various warning messages.
2807 (arc_handle_aux_attribute): Likewise.
2809 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
2811 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
2813 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2815 * doc/sourcebuild.texi: Document usage of line number 0 in verify
2816 compiler messages directives.
2818 2018-06-12 Matthew Fortune <mfortune@gmail.com>
2820 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
2821 * config/mips/mips-tables.opt: Regenerate.
2822 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
2824 * doc/invoke.texi: Document -march=i6500.
2826 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
2828 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
2829 (i6400_gpmul): Add cpu_unit.
2830 (i6400_gpdiv): Likewise.
2831 (i6400_msa_add_d): Update reservations.
2832 (i6400_msa_int_add) Likewise.
2833 (i6400_msa_short_logic3) Likewise.
2834 (i6400_msa_short_logic2) Likewise.
2835 (i6400_msa_short_logic) Likewise.
2836 (i6400_msa_move) Likewise.
2837 (i6400_msa_cmp) Likewise.
2838 (i6400_msa_short_float2) Likewise.
2839 (i6400_msa_div_d) Likewise.
2840 (i6400_msa_long_logic1) Likewise.
2841 (i6400_msa_long_logic2) Likewise.
2842 (i6400_msa_mult) Likewise.
2843 (i6400_msa_long_float2) Likewise.
2844 (i6400_msa_long_float4) Likewise.
2845 (i6400_msa_long_float5) Likewise.
2846 (i6400_msa_long_float8) Likewise.
2847 (i6400_fpu_fadd): Include frint type.
2848 (i6400_fpu_store): New define_insn_reservation.
2849 (i6400_fpu_load): Likewise.
2850 (i6400_fpu_move): Likewise.
2851 (i6400_fpu_fcmp): Likewise.
2852 (i6400_fpu_fmadd): Likewise.
2853 (i6400_int_mult): Include imul3nc type and update reservation.
2854 (i6400_int_div): Include idiv3 type and update reservation.
2855 (i6400_int_load): Update to check type not move_type.
2856 (i6400_int_store): Likewise.
2857 (i6400_int_prefetch): Set zero latency.
2859 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
2861 * gcc.c: Document new %@{...} sequence.
2862 (LINK_COMMAND_SPEC): Use it for the -L switches.
2863 (cpp_unique_options): Use it for the -I switches.
2864 (at_file_argbuf): New global variable.
2865 (in_at_file): Likewise.
2866 (alloc_args): Create at_file_argbuf.
2867 (clear_args): Truncate at_file_argbuf.
2868 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
2869 (open_at_file): New function.
2870 (close_at_file): Likewise.
2871 (create_at_file): Delete.
2872 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
2875 (validate_switches_from_spec): Deal with %@{...} sequence.
2876 (validate_switches): Likewise.
2877 (driver::finalize): Call clear_args.
2879 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2881 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
2883 2018-06-11 Martin Sebor <msebor@redhat.com>
2885 * doc/invoke.texi (-Wall): List -Wc++17-compat.
2886 (Wno-class-memaccess): Add @opindex.
2887 (Wno-templates, Wno-multiple-inheritance): Same.
2888 (Wno-virtual-inheritance, Wno-namespaces): Same.
2889 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
2890 (Wno-format-overflow, Wno-format-truncation): Same.
2891 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
2892 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
2893 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
2894 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
2895 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
2896 (Wno-misspelled-isr): Same.
2898 2018-06-11 Martin Sebor <msebor@redhat.com>
2900 * PR tree-optimization/86083
2901 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
2903 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
2905 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
2907 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
2910 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
2911 on the correct operand.
2912 (*movdi_internal64): Ditto.
2914 2018-06-11 Martin Liska <mliska@suse.cz>
2916 PR tree-optimization/86089
2917 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
2919 2018-06-11 Julia Koval <julia.koval@intel.com>
2921 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
2922 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
2923 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
2925 2018-06-11 Olivier Hainque <hainque@adacore.com>
2927 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
2928 for Ada with strict dwarf2.
2930 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
2933 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
2936 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
2938 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
2940 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
2942 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
2945 2018-06-08 Martin Liska <mliska@suse.cz>
2947 * tree-cfg.h (debug_function): Fix argument type to match
2950 2018-06-08 Martin Liska <mliska@suse.cz>
2952 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
2953 Remove usage of MPX-related (and removed) fields.
2954 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
2956 2018-06-08 David Malcolm <dmalcolm@redhat.com>
2958 * cfg.c (debug): Use TDF_NONE rather than 0.
2959 * cfghooks.c (debug): Likewise.
2960 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
2961 (struct dump_option_value_info): Convert to...
2962 (struct kv_pair): ...this template type.
2963 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
2965 (optinfo_verbosity_options): Likewise.
2966 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
2968 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
2969 than int for "optgroup_flags" param.
2970 (dump_generic_expr_loc): Use dump_flags_t rather than int for
2972 (dump_dec): Likewise.
2973 (dump_finish): Use TDF_NONE rather than 0.
2974 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
2975 rather than int for "optgroup_flags" param. Use TDF_NONE rather
2976 than 0. Update for change to option_ptr.
2977 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
2978 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
2979 0. Update for changes to optinfo_verbosity_options and
2981 (opt_info_switch_p): Convert optgroup_flags from int to
2983 (dump_basic_block): Use dump_flags_t rather than int
2984 for "dump_kind" param.
2985 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
2986 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
2987 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
2988 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
2989 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
2990 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
2991 TDF_NONE): Convert from macros to...
2992 (enum dump_flag): ...this new enum.
2993 (dump_flags_t): Update to use enum.
2994 (operator|, operator&, operator~, operator|=, operator&=):
2995 Implement for dump_flags_t.
2996 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
2997 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
2998 Convert from macros to...
2999 (enum optgroup_flag): ...this new enum.
3000 (optgroup_flags_t): New typedef.
3001 (operator|, operator|=): Implement for optgroup_flags_t.
3002 (struct dump_file_info): Convert field "alt_flags" to
3003 dump_flags_t. Convert field "optgroup_flags" to
3005 (dump_basic_block): Use dump_flags_t rather than int for param.
3006 (dump_generic_expr_loc): Likewise.
3007 (dump_dec): Likewise.
3008 (dump_register): Convert param "optgroup_flags" to
3010 (opt_info_enable_passes): Likewise.
3011 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3013 * gimple-pretty-print.c (debug): Likewise.
3014 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3015 (merged_store_group::apply_stores): Likewise.
3016 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3017 * gimple.c (verify_gimple_pp): Likewise.
3018 * graphite-poly.c (print_pbb_body): Likewise.
3019 * passes.c (pass_manager::register_one_dump_file): Convert
3020 local "optgroup_flags" to optgroup_flags_t.
3021 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3023 (debug_body): Likewise.
3024 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3025 to optgroup_flags_t.
3026 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3028 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3029 (convert_mult_to_fma): Likewise.
3030 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3031 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3032 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3033 "dump_kind" to dump_flags_t.
3035 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3037 * config/rs6000/rs6000.c (min, max): Delete.
3039 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3041 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3042 -mabi=spe and -mabi=no-spe.
3044 2018-06-08 Martin Liska <mliska@suse.cz>
3046 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3047 where we expect an existing summary.
3049 2018-06-08 Martin Liska <mliska@suse.cz>
3051 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3052 * ipa-inline.h (estimate_edge_growth): Likewise.
3054 2018-06-08 Martin Liska <mliska@suse.cz>
3056 * cgraph.c (function_version_hasher::hash): Use
3057 cgraph_node::get_uid ().
3058 (function_version_hasher::equal):
3059 * cgraph.h (cgraph_node::get_uid): New method.
3060 * ipa-inline.c (update_caller_keys): Use
3061 cgraph_node::get_uid ().
3062 (update_callee_keys): Likewise.
3063 * ipa-utils.c (searchc): Likewise.
3064 (ipa_reduced_postorder): Likewise.
3065 * lto-cgraph.c (input_node): Likewise.
3066 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
3067 * symbol-summary.h (symtab_insertion): Likewise.
3068 (symtab_removal): Likewise.
3069 (symtab_duplication): Likewise.
3070 * tree-pretty-print.c (dump_function_header): Likewise.
3071 * tree-sra.c (convert_callers_for_node): Likewise.
3073 2018-06-08 Martin Liska <mliska@suse.cz>
3075 * cgraph.c (symbol_table::create_edge): Always assign a new
3077 (symbol_table::free_edge): Do not recycle numbers.
3078 * cgraph.h (cgraph_edge::get): New method.
3079 * symbol-summary.h (symtab_removal): Use it.
3080 (symtab_duplication): Likewise.
3081 (call_summary::hashable_uid): Remove.
3083 2018-06-08 Martin Liska <mliska@suse.cz>
3085 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
3086 (initialize_growth_caches): Remove.
3087 (free_growth_caches): Likewise.
3088 (do_estimate_edge_time): Use edge_growth_cache.
3089 (do_estimate_edge_size): Likewise.
3090 (do_estimate_edge_hints): Likewise.
3091 * ipa-inline.c (reset_edge_caches): Likewise.
3092 (recursive_inlining): Likewise.
3093 (inline_small_functions): Likewise.
3094 * ipa-inline.h (initialize_growth_caches): Remove.
3095 (estimate_edge_size): Likewise.
3096 (estimate_edge_time): Likewise.
3097 (estimate_edge_hints): Likewise.
3098 (reset_edge_growth_cache): Likewise.
3099 * symbol-summary.h (call_summary::remove): New method.
3101 2018-06-08 Martin Liska <mliska@suse.cz>
3103 * ipa-cp.c (class edge_clone_summary): New summary.
3104 (grow_edge_clone_vectors): Remove.
3105 (ipcp_edge_duplication_hook): Remove.
3106 (class edge_clone_summary_t): New call_summary class.
3107 (ipcp_edge_removal_hook): Remove.
3108 (edge_clone_summary_t::duplicate): New function.
3109 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
3110 (create_specialized_node): Likewise.
3111 (ipcp_driver): Initialize edge_clone_summaries and do not
3114 2018-06-08 Martin Liska <mliska@suse.cz>
3116 * symbol-summary.h (get): New function.
3117 (call_summary::m_initialize_when_cloning): New class member.
3119 2018-06-08 Martin Liska <mliska@suse.cz>
3121 * cgraph.c (cgraph_node::remove): Do not recycle uid.
3122 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
3123 (symbol_table::allocate_cgraph_symbol): Do not set uid.
3124 * passes.c (uid_hash_t): Record removed_nodes by their uids.
3125 (remove_cgraph_node_from_order): Use the removed_nodes set.
3126 (do_per_function_toporder): Likwise.
3127 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
3128 instead of summary_uid.
3129 (symtab_removal): Likewise.
3130 (symtab_duplication): Likewise.
3132 2018-06-08 Martin Liska <mliska@suse.cz>
3134 * ipa-cp.c (ipcp_store_bits_results): Use
3135 ipcp_transformation_sum.
3136 (ipcp_store_vr_results): Likewise.
3137 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
3139 (ipcp_transformation_initialize): ... this.
3140 (ipa_set_node_agg_value_chain):
3141 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
3142 (write_ipcp_transformation_info): Likewise.
3143 (read_ipcp_transformation_info): Likewise.
3144 (ipcp_update_bits): Likewise.
3145 (ipcp_update_vr): Likewise.
3146 (ipcp_transform_function): Likewise.
3147 * ipa-prop.h: Rename ipcp_transformation_summary to
3148 ipcp_transformation.
3149 (class ipcp_transformation_t): New function summary.
3150 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
3151 (ipa_get_agg_replacements_for_node): Likewise.
3153 2018-06-08 Martin Liska <mliska@suse.cz>
3155 * ipa-pure-const.c (struct funct_state_d): Do it class instead
3157 (class funct_state_summary_t): New function_summary class.
3158 (has_function_state): Remove.
3159 (get_function_state): Likewise.
3160 (set_function_state): Likewise.
3161 (add_new_function): Likewise.
3162 (funct_state_summary_t::insert): New function.
3163 (duplicate_node_data): Remove.
3164 (remove_node_data): Remove.
3165 (funct_state_summary_t::duplicate): New function.
3166 (register_hooks): Create new funct_state_summaries.
3167 (pure_const_generate_summary): Use it.
3168 (pure_const_write_summary): Likewise.
3169 (pure_const_read_summary): Likewise.
3170 (propagate_pure_const): Likewise.
3171 (propagate_nothrow): Likewise.
3172 (dump_malloc_lattice): Likewise.
3173 (propagate_malloc): Likewise.
3174 (execute): Do not register hooks, just remove summary
3176 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
3179 2018-06-08 Martin Liska <mliska@suse.cz>
3181 * ipa-reference.c (remove_node_data): Remove.
3182 (duplicate_node_data): Likewise.
3183 (class ipa_ref_var_info_summary_t): New class.
3184 (class ipa_ref_opt_summary_t): Likewise.
3185 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
3186 (get_reference_optimization_summary): Use
3187 ipa_ref_opt_sum_summaries.
3188 (set_reference_vars_info): Remove.
3189 (set_reference_optimization_summary): Likewise.
3190 (ipa_init): Create summaries.
3191 (init_function_info): Use function summary.
3192 (ipa_ref_opt_summary_t::duplicate): New function.
3193 (ipa_ref_opt_summary_t::remove): New function.
3194 (get_read_write_all_from_node): Fix GNU coding style.
3195 (propagate): Use function summary.
3196 (write_node_summary_p): Fix GNU coding style.
3197 (stream_out_bitmap): Likewise.
3198 (ipa_reference_read_optimization_summary): Use function summary.
3199 (ipa_reference_c_finalize): Do not release hooks.
3201 2018-06-08 Martin Liska <mliska@suse.cz>
3203 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
3204 (analyze_function_body): Extract multiple calls of get_create.
3205 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3206 * ipa-inline.c (recursive_inlining): Use ::get method.
3207 * ipa-inline.h (estimate_edge_growth): Likewise.
3209 2018-06-08 Martin Liska <mliska@suse.cz>
3211 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
3213 (hsa_function_summary::hsa_function_summary): Use the new enum
3215 (hsa_gpu_implementation_p): Use hsa_summaries::get.
3216 * hsa-gen.c (hsa_get_host_function): Likewise.
3217 (get_brig_function_name): Likewise.
3218 * ipa-hsa.c (process_hsa_functions): Likewise.
3219 (ipa_hsa_write_summary): Likewise.
3220 * symbol-summary.h (symtab_duplication): Use ::get function/
3221 (get): New function.
3223 2018-06-08 Martin Liska <mliska@suse.cz>
3225 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
3227 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
3228 (hsa_register_kernel): Likewise.
3229 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
3230 * hsa-gen.c (hsa_get_host_function): Likewise.
3231 (get_brig_function_name): Likewise.
3232 (generate_hsa): Likewise.
3233 (pass_gen_hsail::execute): Likewise.
3234 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3235 (devirtualization_time_bonus): Likewise.
3236 (ipcp_propagate_stage): Likewise.
3237 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3238 (edge_set_predicate): Likewise.
3239 (evaluate_conditions_for_known_args): Likewise.
3240 (evaluate_properties_for_edge): Likewise.
3241 (ipa_fn_summary::reset): Likewise.
3242 (ipa_fn_summary_t::duplicate): Likewise.
3243 (dump_ipa_call_summary): Likewise.
3244 (ipa_dump_fn_summary): Likewise.
3245 (analyze_function_body): Likewise.
3246 (compute_fn_summary): Likewise.
3247 (estimate_edge_devirt_benefit): Likewise.
3248 (estimate_edge_size_and_time): Likewise.
3249 (estimate_calls_size_and_time): Likewise.
3250 (estimate_node_size_and_time): Likewise.
3251 (inline_update_callee_summaries): Likewise.
3252 (remap_edge_change_prob): Likewise.
3253 (remap_edge_summaries): Likewise.
3254 (ipa_merge_fn_summary_after_inlining): Likewise.
3255 (ipa_update_overall_fn_summary): Likewise.
3256 (read_ipa_call_summary): Likewise.
3257 (inline_read_section): Likewise.
3258 (write_ipa_call_summary): Likewise.
3259 (ipa_fn_summary_write): Likewise.
3260 (ipa_free_fn_summary): Likewise.
3261 * ipa-hsa.c (process_hsa_functions): Likewise.
3262 (ipa_hsa_write_summary): Likewise.
3263 (ipa_hsa_read_section): Likewise.
3264 * ipa-icf.c (sem_function::merge): Likewise.
3265 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
3266 (do_estimate_edge_time): Likewise.
3267 (estimate_size_after_inlining): Likewise.
3268 (estimate_growth): Likewise.
3269 (growth_likely_positive): Likewise.
3270 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3271 (inline_call): Likewise.
3272 * ipa-inline.c (caller_growth_limits): Likewise.
3273 (can_inline_edge_p): Likewise.
3274 (can_inline_edge_by_limits_p): Likewise.
3275 (compute_uninlined_call_time): Likewise.
3276 (compute_inlined_call_time): Likewise.
3277 (want_inline_small_function_p): Likewise.
3278 (edge_badness): Likewise.
3279 (update_caller_keys): Likewise.
3280 (update_callee_keys): Likewise.
3281 (recursive_inlining): Likewise.
3282 (inline_small_functions): Likewise.
3283 (inline_to_all_callers_1): Likewise.
3284 (dump_overall_stats): Likewise.
3285 (early_inline_small_functions): Likewise.
3286 (early_inliner): Likewise.
3287 * ipa-inline.h (estimate_edge_growth): Likewise.
3288 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3289 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3290 * ipa-prop.h (IPA_NODE_REF): Likewise.
3291 (IPA_EDGE_REF): Likewise.
3292 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3293 (propagate_malloc): Likewise.
3294 * ipa-split.c (execute_split_functions): Likewise.
3295 * symbol-summary.h: Rename get to get_create.
3297 (get_create): Likewise.
3298 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3300 2018-06-08 Martin Liska <mliska@suse.cz>
3302 * symbol-summary.h (release): Move definition out of class
3304 (symtab_removal): Likewise.
3305 (symtab_duplication): Likewise.
3307 2018-06-08 Martin Liska <mliska@suse.cz>
3309 * symbol-summary.h (function_summary): Move constructor
3310 implementation out of class declaration.
3311 (release): Likewise.
3312 (symtab_insertion): Likewise.
3313 (symtab_removal): Likewise.
3314 (symtab_duplication): Likewise.
3317 2018-06-08 Martin Liska <mliska@suse.cz>
3319 * Makefile.in: Remove support for MPX (macros, related functions,
3320 fields in cgraph_node, ...).
3321 * builtin-types.def (BT_BND): Likewise.
3322 (BT_FN_BND_CONST_PTR): Likewise.
3323 (BT_FN_CONST_PTR_BND): Likewise.
3324 (BT_FN_VOID_PTR_BND): Likewise.
3325 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
3326 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
3327 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
3328 (expand_builtin_mempcpy_with_bounds): Likewise.
3329 (expand_builtin_memset_with_bounds): Likewise.
3330 (expand_builtin_memset_args): Likewise.
3331 (std_expand_builtin_va_start): Likewise.
3332 (expand_builtin): Likewise.
3333 (expand_builtin_with_bounds): Likewise.
3334 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
3335 (DEF_LIB_BUILTIN_CHKP): Likewise.
3336 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
3337 (DEF_CHKP_BUILTIN): Likewise.
3338 (BUILT_IN_MEMCPY): Likewise.
3339 (BUILT_IN_MEMMOVE): Likewise.
3340 (BUILT_IN_MEMPCPY): Likewise.
3341 (BUILT_IN_MEMSET): Likewise.
3342 (BUILT_IN_STPCPY): Likewise.
3343 (BUILT_IN_STRCAT): Likewise.
3344 (BUILT_IN_STRCHR): Likewise.
3345 (BUILT_IN_STRCPY): Likewise.
3346 (BUILT_IN_STRLEN): Likewise.
3347 (BUILT_IN_MEMCPY_CHK): Likewise.
3348 (BUILT_IN_MEMMOVE_CHK): Likewise.
3349 (BUILT_IN_MEMPCPY_CHK): Likewise.
3350 (BUILT_IN_MEMSET_CHK): Likewise.
3351 (BUILT_IN_STPCPY_CHK): Likewise.
3352 (BUILT_IN_STRCAT_CHK): Likewise.
3353 (BUILT_IN_STRCPY_CHK): Likewise.
3354 * calls.c (store_bounds): Likewise.
3355 (emit_call_1): Likewise.
3356 (special_function_p): Likewise.
3357 (maybe_warn_nonstring_arg): Likewise.
3358 (initialize_argument_information): Likewise.
3359 (finalize_must_preallocate): Likewise.
3360 (compute_argument_addresses): Likewise.
3361 (expand_call): Likewise.
3362 * cfgexpand.c (expand_call_stmt): Likewise.
3363 (expand_return): Likewise.
3364 (expand_gimple_stmt_1): Likewise.
3365 (pass_expand::execute): Likewise.
3366 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3367 (cgraph_node::remove): Likewise.
3368 (cgraph_node::dump): Likewise.
3369 (cgraph_node::verify_node): Likewise.
3370 * cgraph.h (chkp_function_instrumented_p): Likewise.
3371 (symtab_node::get_alias_target): Likewise.
3372 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
3373 (cgraph_local_p): Likewise.
3374 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
3375 (cgraph_edge::rebuild_references): Likewise.
3376 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
3377 (walk_polymorphic_call_targets): Likewise.
3378 (cgraph_node::expand_thunk): Likewise.
3379 (symbol_table::output_weakrefs): Likewise.
3380 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
3381 (ix86_handle_option): Likewise.
3382 * config/i386/constraints.md: Likewise.
3383 * config/i386/i386-builtin-types.def (BND): Likewise.
3387 * config/i386/i386-builtin.def (BDESC_END): Likewise.
3388 (BDESC_FIRST): Likewise.
3390 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
3391 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
3392 * config/i386/i386.c (enum reg_class): Likewise.
3393 (ix86_target_string): Likewise.
3394 (ix86_option_override_internal): Likewise.
3395 (ix86_conditional_register_usage): Likewise.
3396 (ix86_valid_target_attribute_inner_p): Likewise.
3397 (ix86_set_indirect_branch_type): Likewise.
3398 (ix86_set_current_function): Likewise.
3399 (ix86_function_arg_regno_p): Likewise.
3400 (init_cumulative_args): Likewise.
3401 (ix86_function_arg_advance): Likewise.
3402 (ix86_function_arg): Likewise.
3403 (ix86_pass_by_reference): Likewise.
3404 (ix86_function_value_regno_p): Likewise.
3405 (ix86_function_value_1): Likewise.
3406 (ix86_function_value_bounds): Likewise.
3407 (ix86_return_in_memory): Likewise.
3408 (ix86_setup_incoming_vararg_bounds): Likewise.
3409 (ix86_va_start): Likewise.
3410 (indirect_thunk_need_prefix): Likewise.
3411 (print_reg): Likewise.
3412 (ix86_print_operand): Likewise.
3413 (ix86_expand_call): Likewise.
3414 (ix86_output_function_return): Likewise.
3415 (reg_encoded_number): Likewise.
3416 (BDESC_VERIFYS): Likewise.
3417 (ix86_init_mpx_builtins): Likewise.
3418 (ix86_init_builtins): Likewise.
3419 (ix86_emit_cmove): Likewise.
3420 (ix86_emit_move_max): Likewise.
3421 (ix86_expand_builtin): Likewise.
3422 (ix86_builtin_mpx_function): Likewise.
3423 (ix86_get_arg_address_for_bt): Likewise.
3424 (ix86_load_bounds): Likewise.
3425 (ix86_store_bounds): Likewise.
3426 (ix86_load_returned_bounds): Likewise.
3427 (ix86_store_returned_bounds): Likewise.
3428 (ix86_class_likely_spilled_p): Likewise.
3429 (ix86_hard_regno_mode_ok): Likewise.
3430 (x86_order_regs_for_local_alloc): Likewise.
3431 (ix86_mitigate_rop): Likewise.
3432 (ix86_bnd_prefixed_insn_p): Likewise.
3433 (ix86_mpx_bound_mode): Likewise.
3434 (ix86_make_bounds_constant): Likewise.
3435 (ix86_initialize_bounds): Likewise.
3436 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
3437 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
3438 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
3439 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
3440 (TARGET_CHKP_BOUND_MODE): Likewise.
3441 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
3442 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
3443 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
3444 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
3445 * config/i386/i386.h (TARGET_MPX): Likewise.
3446 (TARGET_MPX_P): Likewise.
3447 (VALID_BND_REG_MODE): Likewise.
3448 (FIRST_BND_REG): Likewise.
3449 (LAST_BND_REG): Likewise.
3450 (enum reg_class): Likewise.
3451 (BND_REG_P): Likewise.
3452 (BND_REGNO_P): Likewise.
3453 (BNDmode): Likewise.
3454 (ADJUST_INSN_LENGTH): Likewise.
3455 * config/i386/i386.md: Likewise.
3456 * config/i386/i386.opt: Likewise.
3457 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
3458 (defined): Likewise.
3459 (LINK_MPX): Likewise.
3460 (MPX_SPEC): Likewise.
3461 (LIBMPX_SPEC): Likewise.
3462 (LIBMPXWRAPPERS_SPEC): Likewise.
3463 (CHKP_SPEC): Likewise.
3464 * config/i386/predicates.md: Likewise.
3465 * dbxout.c (dbxout_type): Likewise.
3466 * doc/extend.texi: Likewise.
3467 * doc/invoke.texi: Likewise.
3468 * doc/md.texi: Likewise.
3469 * doc/tm.texi: Likewise.
3470 * doc/tm.texi.in: Likewise.
3471 * dwarf2out.c (is_base_type): Likewise.
3472 (gen_formal_types_die): Likewise.
3473 (gen_subprogram_die): Likewise.
3474 (gen_type_die_with_usage): Likewise.
3475 (gen_decl_die): Likewise.
3476 (dwarf2out_late_global_decl): Likewise.
3477 * expr.c (expand_assignment): Likewise.
3478 (emit_storent_insn): Likewise.
3479 (store_expr_with_bounds): Likewise.
3480 (store_expr): Likewise.
3481 (expand_expr_real_1): Likewise.
3482 * expr.h (store_expr_with_bounds): Likewise.
3483 * function.c (use_register_for_decl): Likewise.
3484 (struct bounds_parm_data): Likewise.
3485 (assign_parms_augmented_arg_list): Likewise.
3486 (assign_parm_find_entry_rtl): Likewise.
3487 (assign_parm_is_stack_parm): Likewise.
3488 (assign_parm_load_bounds): Likewise.
3489 (assign_bounds): Likewise.
3490 (assign_parms): Likewise.
3491 (expand_function_start): Likewise.
3492 * gcc.c (CHKP_SPEC): Likewise.
3493 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
3494 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
3495 (wrestrict_dom_walker::check_call): Likewise.
3496 * gimple.c (gimple_build_call_from_tree): Likewise.
3497 * gimple.h (enum gf_mask): Likewise.
3498 (gimple_call_with_bounds_p): Likewise.
3499 (gimple_call_set_with_bounds): Likewise.
3500 * gimplify.c (gimplify_init_constructor): Likewise.
3501 * ipa-cp.c (initialize_node_lattices): Likewise.
3502 (propagate_constants_across_call): Likewise.
3503 (find_more_scalar_values_for_callers_subset): Likewise.
3504 * ipa-hsa.c (process_hsa_functions): Likewise.
3505 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
3506 * ipa-icf.c (sem_function::merge): Likewise.
3507 * ipa-inline.c (early_inliner): Likewise.
3508 * ipa-pure-const.c (warn_function_noreturn): Likewise.
3509 (warn_function_cold): Likewise.
3510 (propagate_pure_const): Likewise.
3511 * ipa-ref.h (enum GTY): Likewise.
3512 * ipa-split.c (find_retbnd): Likewise.
3513 (consider_split): Likewise.
3514 (split_function): Likewise.
3515 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3516 * ipa.c (walk_polymorphic_call_targets): Likewise.
3517 (symbol_table::remove_unreachable_nodes): Likewise.
3518 (process_references): Likewise.
3519 (cgraph_build_static_cdtor_1): Likewise.
3520 * lto-cgraph.c (lto_output_node): Likewise.
3521 (output_refs): Likewise.
3522 (compute_ltrans_boundary): Likewise.
3523 (input_overwrite_node): Likewise.
3524 (input_node): Likewise.
3525 (input_cgraph_1): Likewise.
3526 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
3527 * passes.c (pass_manager::execute_early_local_passes): Likewise.
3528 (class pass_chkp_instrumentation_passes): Likewise.
3529 (make_pass_chkp_instrumentation_passes): Likewise.
3530 * passes.def: Likewise.
3531 * rtl.h (struct GTY): Likewise.
3532 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
3533 * stor-layout.c (layout_type): Likewise.
3534 * symtab.c: Likewise.
3535 * target.def: Likewise.
3536 * targhooks.c (default_chkp_bound_type): Likewise.
3537 (default_chkp_bound_mode): Likewise.
3538 (default_builtin_chkp_function): Likewise.
3539 (default_chkp_function_value_bounds): Likewise.
3540 (default_chkp_make_bounds_constant): Likewise.
3541 (default_chkp_initialize_bounds): Likewise.
3542 * targhooks.h (default_chkp_bound_type): Likewise.
3543 (default_chkp_bound_mode): Likewise.
3544 (default_builtin_chkp_function): Likewise.
3545 (default_chkp_function_value_bounds): Likewise.
3546 (default_chkp_make_bounds_constant): Likewise.
3547 (default_chkp_initialize_bounds): Likewise.
3548 * toplev.c (compile_file): Likewise.
3549 (process_options): Likewise.
3550 * tree-core.h (DEF_BUILTIN): Likewise.
3551 (DEF_BUILTIN_CHKP): Likewise.
3552 * tree-inline.c (declare_return_variable): Likewise.
3553 (remap_gimple_stmt): Likewise.
3554 (copy_bb): Likewise.
3555 (initialize_inlined_parameters): Likewise.
3556 (expand_call_inline): Likewise.
3557 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
3558 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
3559 (make_pass_ipa_chkp_produce_thunks): Likewise.
3560 (make_pass_chkp): Likewise.
3561 (make_pass_chkp_opt): Likewise.
3562 (make_pass_chkp_instrumentation_passes): Likewise.
3563 * tree-pretty-print.c (dump_generic_node): Likewise.
3564 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
3565 * tree-ssa-dce.c (propagate_necessity): Likewise.
3566 (eliminate_unnecessary_stmts): Likewise.
3567 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
3568 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3569 * tree-ssa-sccvn.h: Likewise.
3570 * tree-ssa-strlen.c (get_string_length): Likewise.
3571 (valid_builtin_call): Likewise.
3572 (adjust_last_stmt): Likewise.
3573 (handle_builtin_strchr): Likewise.
3574 (handle_builtin_strcpy): Likewise.
3575 (handle_builtin_stxncpy): Likewise.
3576 (handle_builtin_memcpy): Likewise.
3577 (handle_builtin_strcat): Likewise.
3578 (strlen_check_and_optimize_stmt): Likewise.
3579 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
3580 * tree-streamer-in.c: Likewise.
3581 * tree-streamer.c (record_common_node): Likewise.
3582 * tree.c (tree_code_size): Likewise.
3583 (wide_int_to_tree_1): Likewise.
3584 (type_contains_placeholder_1): Likewise.
3585 (build_common_tree_nodes): Likewise.
3586 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
3587 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
3588 (POINTER_BOUNDS_P): Likewise.
3589 (BOUNDED_TYPE_P): Likewise.
3590 (BOUNDED_P): Likewise.
3591 (CALL_WITH_BOUNDS_P): Likewise.
3592 (pointer_bounds_type_node): Likewise.
3593 * value-prof.c (gimple_ic): Likewise.
3594 * var-tracking.c (vt_add_function_parameters): Likewise.
3595 * varasm.c (make_decl_rtl): Likewise.
3596 (assemble_start_function): Likewise.
3597 (output_constant): Likewise.
3598 (maybe_assemble_visibility): Likewise.
3599 * varpool.c (ctor_for_folding): Likewise.
3600 * chkp-builtins.def: Remove.
3601 * ipa-chkp.c: Remove.
3602 * ipa-chkp.h: Remove.
3603 * rtl-chkp.c: Remove.
3604 * rtl-chkp.h: Remove.
3605 * tree-chkp-opt.c: Remove.
3606 * tree-chkp.c: Remove.
3607 * tree-chkp.h: Remove.
3609 2018-06-07 Carl Love <cel@us.ibm.com>
3611 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3612 vextract_fp_from_shortl): Add BE support.
3614 2018-06-07 Paul Koning <ni1d@arrl.net>
3616 * compare-elim.c (try_merge_compare): Don't merge compare if
3617 address contains a side effect.
3618 (try_eliminate_compare): Likewise.
3620 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
3622 * config.gcc: Support "tremont".
3623 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
3624 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3626 * config/i386/i386.c (m_TREMONT): Define.
3627 (processor_target_table): Add "tremont".
3628 (PTA_TREMONT): Define.
3629 (ix86_lea_outperforms): Add TARGET_TREMONT.
3630 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
3631 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
3632 and M_INTEL_GOLDMONT_PLUS.
3633 (fold_builtin_cpu): Add "tremont".
3634 (ix86_add_stmt_cost): Add TARGET_TREMONT.
3635 (ix86_option_override_internal): Add "tremont".
3636 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
3637 (processor_type): Add PROCESSOR_TREMONT.
3638 * config/i386/x86-tune.def: Add m_TREMONT.
3639 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
3641 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3643 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
3644 symbol defined for msp430i* devices to be lower case.
3646 gcc/testsuite/gcc.target/msp430/
3647 * msp430i-device-symbol.c: New test.
3648 * msp430f-device-symbol.c: New test.
3649 * msp430.h: New test header file.
3651 2018-06-07 Richard Biener <rguenther@suse.de>
3653 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
3654 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
3655 Properly wrap signed arithmetic if overflow wraps.
3657 2018-06-07 Jakub Jelinek <jakub@redhat.com>
3659 PR tree-optimization/69615
3660 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
3661 of a cast from a same precision integral SSA_NAME in a bb dominated
3662 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
3663 cast to utype if rhs2 has already a compatible type.
3665 2018-06-07 Richard Biener <rguenther@suse.de>
3667 PR tree-optimization/85935
3668 * graphite-scop-detection.c (find_params_in_bb): Analyze
3669 condition operands with respect to the correct loop. Assert
3670 the analysis doesn't fail.
3672 2018-06-04 Carl Love <cel@us.ibm.com>
3674 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
3675 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
3676 as it is slightly cheaper.
3677 (first_match_or_eos_index_<mode>):
3678 Calculate index using natural element order.
3679 (first_match_index_<mode>):
3680 Calculate index using natural element order.
3681 (first_match_or_eos_index_<mode>):
3682 Calculate index using natural order.
3683 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3684 for BE and LE modes.
3685 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3686 P9V_BUILTIN_VCLZLSBB_V16QI.
3687 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3690 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3692 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
3693 indentation and line wrap for many prototypes. Add missing
3694 @smallexample directives around block of prototypes for vec_xl and
3697 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
3699 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3700 track if we pass or return IEEE 128-bit floating point.
3701 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3702 C++ mangling that is compatible with GCC 8.1.
3703 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3704 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3705 floating point types.
3706 (rs6000_function_arg_advance_1): Likewise.
3707 (rs6000_mangle_type): Optionally generate mangled names that match
3708 what GCC 8.1 generated for IEEE 128-bit floating point types.
3709 (rs6000_globalize_decl_name): If we have an external function that
3710 passes or returns IEEE 128-bit types, generate a weak reference
3711 from the mangled name used in GCC 8.1 to the current mangled
3713 (rs6000_init_builtins): Make __ibm128 use the long double type if
3714 long double is IBM extended double. Make __float128 use the long
3715 double type if long double is IEEE 128-bit.
3718 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3719 macro for __ibm128 built-in functions.
3720 (PACK_IF): Add __ibm128 pack/unpack functions.
3721 (UNPACK_IF): Likewise.
3722 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3723 enable long double built-in functions if long double is IEEE
3724 128-bit floating point.
3725 (rs6000_invalid_builtin): Update long double built-in function
3727 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3728 functions, adjust the built-in function to use the long double
3729 built-in function if __ibm128 and long double are the same type.
3730 * doc/extend.texi (PowerPC builtins): Update documention for
3731 __builtin_{,un}pack_longdouble. Add documentation for
3732 __builtin_{,un}pack_ibm128.
3734 2018-06-06 Jim Wilson <jimw@sifive.com>
3736 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
3737 (struct machine_function): New field interrupt_mode.
3738 (riscv_handle_type_attribute): New function. Add forward declaration.
3739 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
3740 (riscv_expand_epilogue): Check interrupt_mode field.
3741 (riscv_set_current_function): Check interrupt attribute args and
3742 set interrupt_mode field.
3743 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
3744 (riscv_sret, riscv_uret): New.
3745 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
3746 new arguments to interrupt attribute.
3748 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3751 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3752 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3754 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
3756 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
3757 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
3758 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3759 several redundant entries.
3761 2018-06-06 David Malcolm <dmalcolm@redhat.com>
3763 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
3764 type from "rtx" to "rtx_insn *".
3765 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
3766 for local "call_insn", removing cast.
3767 (ix86_expand_call): Likewise, introducing a "call_insn" local.
3769 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
3771 PR tree-optimization/86066
3772 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
3773 for BIT_INSERT_EXPR stores.
3775 2018-06-06 Richard Biener <rguenther@suse.de>
3777 PR tree-optimization/86062
3778 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3779 component refs ontop
3780 of to be offsetted base.
3782 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3784 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
3785 to be static and remove check on interrupt attribute name.
3787 gcc/testsuite/gcc.target/msp430/
3788 * function-attributes-4.c: New test.
3789 * static-interrupts.c: New test.
3791 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3793 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3794 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
3796 2018-06-05 Steve Ellcey <sellcey@cavium.com>
3799 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
3801 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
3802 Remove second argument, change how error is called.
3803 (aarch64_layout_arg): Remove second argument from
3804 aarch64_err_no_fpadvsimd call.
3805 (aarch64_init_cumulative_args): Ditto.
3806 (aarch64_gimplify_va_arg_expr): Ditto.
3807 * config/aarch64/aarch64.md (mov<mode>): Ditto.
3809 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3811 * config/i386/i386.md (simple_return_indirect_internal): New expander.
3812 (*simple_return_indirect_internal<mode>): Rename from
3813 simple_return_indirect_internal. Use W mode iterator.
3814 (rstorssp): New expander.
3815 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
3816 (clrssbsy): New expander.
3817 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
3819 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
3821 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
3823 (cmse_check_pointed_object): Likewise.
3825 2018-06-05 Martin Liska <mliska@suse.cz>
3827 PR gcov-profile/47618
3828 * doc/invoke.texi: Document how -fprofile-dir format
3831 2018-06-05 Richard Biener <rguenther@suse.de>
3833 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
3834 removal pretend DOM info isn't available so we do not update
3835 it and only remove edges, not dominated blocks. Actually free
3836 DOM info in case we removed something. Remove unreachable blocks.
3837 (mfb_keep_latches): Work with either DOM info or marked backedges.
3838 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
3839 first. Mark backedges if DOM info isn't available.
3840 (Re-)compute DOM info after cleanup_control_flow_pre.
3842 2018-06-05 Richard Biener <rguenther@suse.de>
3844 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
3845 (locus_discrim_hasher::hash): Adjust.
3846 (locus_discrim_hasher::equal): Likewise.
3847 (next_discriminator_for_locus): Work on line directly.
3848 (same_line_p): Pass in expanded locus1 as well.
3849 (assign_discriminators): Avoid redundant location expansions.
3851 2018-06-05 Richard Biener <rguenther@suse.de>
3853 PR tree-optimization/86046
3854 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
3855 if required after clearing TREE_ADDRESSABLE.
3857 2018-06-05 Richard Biener <rguenther@suse.de>
3859 PR tree-optimization/86047
3860 * tree-ssa-loop.c (for_each_index): Glob handling of all
3861 decls and constants and really handle all of them.
3863 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3866 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3867 qualifier_void_pointer and qualifier_const_void_pointer.
3868 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3869 (arm_init_builtins): Handle the above.
3870 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3871 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3872 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3875 2018-06-05 Martin Liska <mliska@suse.cz>
3877 * auto-profile.c (read_autofdo_file): Do not use
3878 gcov_ctr_summary struct.
3879 (afdo_callsite_hot_enough_for_early_inline): Likewise.
3880 * coverage.c (struct counts_entry): Likewise.
3881 (read_counts_file): Read just single summary entry.
3882 (get_coverage_counts): Use gcov_summary struct.
3883 * coverage.h (get_coverage_counts): Likewise.
3884 * gcov-dump.c (dump_working_sets): Likewise.
3885 (tag_summary): Dump just single summary.
3886 * gcov-io.c (gcov_write_summary): Write just histogram
3888 (gcov_read_summary): Read just single summary.
3889 (compute_working_sets): Use gcov_summary struct.
3890 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
3891 of GCOV_COUNTERS_SUMMABLE.
3892 (GCOV_COUNTERS_SUMMABLE): Remove.
3893 (GCOV_FIRST_VALUE_COUNTER): Replace with
3894 GCOV_COUNTER_V_INTERVAL.
3895 (struct gcov_ctr_summary): Remove.
3896 (struct gcov_summary): Directly use fields of former
3898 (compute_working_sets): Use gcov_summary struct.
3899 * gcov.c (read_count_file): Do not use ctrs fields.
3900 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
3902 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
3904 * profile.c: Likewise.
3905 * profile.h: Likewise.
3907 2018-06-05 Martin Liska <mliska@suse.cz>
3909 PR gcov-profile/84846
3910 * gcov.c (output_lines): Print working directory only
3911 in intermediate format.
3913 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3915 * config/s390/s390-builtin-types.def: Add void function type.
3916 * config/s390/s390-builtins.def: Use the function type for the
3919 2018-06-04 Jim Wilson <jimw@sifive.com>
3921 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
3923 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
3924 and maybe_eh_return. Change regno to unsigned int. Use new args to
3925 handle EH_RETURN_DATA_REGNO registers properly.
3926 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
3927 (riscv_expand_epilogue): Update comment. Change argument name and
3928 type. Update code to use new name and type. Pass new args to
3929 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
3931 * config/riscv/riscv.md (NORMAL_RETURN): New.
3932 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
3933 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
3934 (eh_return): Call gen_eh_return_internal and emit barrier.
3935 (eh_return_internal): Call riscv_expand_epilogue.
3937 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3939 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
3940 bit_insertion field and declare can_be_merged_into method.
3941 (merged_store_group::can_be_merged_into): New method.
3942 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
3943 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
3944 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
3946 2018-06-04 Richard Biener <rguenther@suse.de>
3948 PR tree-optimization/85955
3949 * builtins.c (fold_builtin_sincos): Convert pointers to
3950 destination to appropriate type before dereferencing.
3952 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3954 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
3956 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
3958 * expr.c (expand_expr_real_1): Force the operand into memory if
3959 its TYPE_MODE is BLKmode and if there is no integer mode for
3960 the number of bits being extracted.
3962 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3966 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3967 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
3969 2018-06-04 Richard Biener <rguenther@suse.de>
3971 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
3972 (cleanup_tree_cfg_noloop): ... single caller. Do
3973 start_recording_case_labels later.
3975 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
3977 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
3978 to _IMMINTRIN_H_INCLUDED.
3979 * config/i386/pconfigintrin.h: Ditto.
3980 * config/i386/waitpkgintrin.h: Ditto.
3981 * config/i386/immintrin.h: Add includes for sgxintrin.h,
3982 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
3983 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
3984 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
3985 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
3986 waitpkgintrin.h and cldemoteintrin.h.
3988 2018-06-04 Richard Biener <rguenther@suse.de>
3990 PR tree-optimization/86038
3991 * tracer.c (find_best_successor): Check probability for
3992 being initialized, bail out if not.
3994 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3997 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3998 of bits to ignore when comparing architectures.
4000 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4002 PR tree-optimization/69615
4003 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4004 maximum or minimum of the type, try to merge it also as if
4005 range1 is + [-, x - 1] or + [x + 1, -].
4008 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4010 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4012 PR tree-optimization/86034
4013 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4014 the unsigned bitfield type in a bit insertion sequence if it does not
4015 have a larger precision than the bitfield size.
4016 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4018 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4020 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4022 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4024 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4025 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4026 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4027 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4029 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4031 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4032 Disable -fdelete-null-pointer-checks for ELF toolchain.
4034 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4035 Kito Cheng <kito.cheng@gmail.com>
4037 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4038 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4039 (nds32*-*-*): Add glibc and uclibc conditions.
4040 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4041 (TARGET_EXCEPT_UNWIND_INFO): Define.
4042 * config/nds32/elf.h: New file.
4043 * config/nds32/linux.h: New file.
4044 * config/nds32/nds32-elf.opt: New file.
4045 * config/nds32/nds32-linux.opt: New file.
4046 * config/nds32/nds32-fp-as-gp.c
4047 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4048 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4050 (nds32_asm_file_end): Ditto.
4051 (nds32_print_operand): Ditto.
4052 (nds32_insert_attributes): Ditto.
4053 (nds32_init_libfuncs): New function.
4054 (TARGET_HAVE_TLS): Define.
4055 (TARGET_INIT_LIBFUNCS): Define.
4056 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4058 (TARGET_ELF): Apply different mcmodel setting.
4059 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4060 been migrated into elf.h and linux.h files.
4061 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4062 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
4063 (mcmodel): The content has been migrated into nds32-elf.opt and
4064 nds32-linux.opt files.
4065 * config/nds32/t-elf: New file.
4066 * config/nds32/t-linux: New file.
4068 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4069 Shiva Chen <shiva0217@gmail.com>
4071 * config/nds32/constants.md (unspec_volatile_element): Add
4072 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
4073 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
4075 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
4076 (make_pass_nds32_fp_as_gp): Declare.
4077 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
4079 (nds32_asm_function_end_prologue): Remove unused asm output.
4080 (nds32_asm_function_begin_epilogue): Remove unused asm output.
4081 (nds32_asm_file_start): Output necessary fp_as_gp information.
4082 (nds32_option_override): Adjust register usage.
4083 (nds32_expand_prologue): Consider fp_as_gp situation.
4084 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
4085 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
4088 (simple_return): Ditto.
4089 (omit_fp_begin): Output special directive for fp_as_gp.
4090 (omit_fp_end): Output special directive for fp_as_gp.
4091 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
4092 mforbid-fp-as-gp): New options.
4094 2018-06-01 Mark Wielaard <mark@klomp.org>
4096 * dwarf2out.c (dwarf2out_finish): Remove generation of
4097 DW_AT_loclists_base.
4099 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
4101 * gimple-ssa-store-merging.c: Include gimple-fold.h.
4102 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
4103 (struct merged_store_group): Add bit_insertion field.
4104 (dump_char_array): Use standard hexadecimal format.
4105 (merged_store_group::merged_store_group): Set bit_insertion to false.
4106 (merged_store_group::apply_stores): Use optimal buffer size. Deal
4107 with BIT_INSERT_EXPR stores. Move up code updating the mask and
4108 also print the mask in the dump file.
4109 (pass_store_merging::gate): Minor tweak.
4110 (imm_store_chain_info::coalesce_immediate): Fix wrong association
4111 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
4112 stores with INTEGER_CST stores.
4113 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
4114 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
4115 and use it throughout. Generate bit insertion sequences if need be.
4116 (pass_store_merging::process_store): Remove redundant condition.
4117 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
4119 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
4121 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
4122 the 128-bit floating point types. Fix function comment.
4124 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4126 * config/aarch64/aarch64-simd.md
4127 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
4129 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
4132 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4134 PR tree-optimization/85989
4135 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
4137 (backprop::intersect_uses): Check it when deciding whether this
4138 is a backedge reference.
4139 (backprop::process_block): Add each phi to m_visited_phis
4140 after visiting it, then clear it at the end.
4142 2018-06-01 Richard Biener <rguenther@suse.de>
4144 * tree-vectorizer.h (vect_dr_stmt): New function.
4145 (vect_get_load_cost): Adjust.
4146 (vect_get_store_cost): Likewise.
4147 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
4148 Use vect_dr_stmt instead of DR_SMTT.
4149 (vect_record_base_alignments): Likewise.
4150 (vect_calculate_target_alignment): Likewise.
4151 (vect_compute_data_ref_alignment): Likewise and make static.
4152 (vect_update_misalignment_for_peel): Likewise.
4153 (vect_verify_datarefs_alignment): Likewise.
4154 (vector_alignment_reachable_p): Likewise.
4155 (vect_get_data_access_cost): Likewise. Pass down
4156 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
4157 (vect_get_peeling_costs_all_drs): Likewise.
4158 (vect_peeling_hash_get_lowest_cost): Likewise.
4159 (vect_enhance_data_refs_alignment): Likewise.
4160 (vect_find_same_alignment_drs): Likewise.
4161 (vect_analyze_data_refs_alignment): Likewise.
4162 (vect_analyze_group_access_1): Likewise.
4163 (vect_analyze_group_access): Likewise.
4164 (vect_analyze_data_ref_access): Likewise.
4165 (vect_analyze_data_ref_accesses): Likewise.
4166 (vect_vfa_segment_size): Likewise.
4167 (vect_small_gap_p): Likewise.
4168 (vectorizable_with_step_bound_p): Likewise.
4169 (vect_prune_runtime_alias_test_list): Likewise.
4170 (vect_analyze_data_refs): Likewise.
4171 (vect_supportable_dr_alignment): Likewise.
4172 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4173 (vect_gen_prolog_loop_niters): Likewise.
4174 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4175 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
4177 (vect_recog_mask_conversion_pattern): Likewise.
4178 (vect_try_gather_scatter_pattern): Likewise.
4179 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
4180 to vect_get_store_cost.
4181 (vect_get_store_cost): Get stmt_info instead of DR.
4182 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
4183 (vect_get_load_cost): Get stmt_info instead of DR.
4185 2018-06-01 Richard Biener <rguenther@suse.de>
4188 * gimple-fold.c (var_decl_component_p): Also allow offsetted
4189 vars wrapped in MEM_REFs.
4191 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
4193 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4194 Fix subreg tests so that we only return a choice between
4195 GENERAL_REGS and FP_REGS if the original classes included both.
4197 2018-06-01 Richard Biener <rguenther@suse.de>
4200 * tree-ssa-structalias.c (get_function_part_constraint):
4201 Handle NULL fi->decl.
4202 (find_func_aliases_for_call): Properly handle indirect
4203 fi from direct call.
4204 (find_func_clobbers): Likewise.
4205 (ipa_pta_execute): Likewise.
4206 (create_variable_info_for): For functions that are ifunc_resolver
4207 resolve to a varinfo that contains the result of the resolver call.
4208 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
4211 2018-05-31 Michael Collison <michael.collison@arm.com>
4213 * config/aarch64/aarch64.md:
4214 (*fix_to_zero_extenddfdi2): New pattern.
4215 * gcc.target/aarch64/fix_extend1.c: New testcase.
4217 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
4221 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
4222 and BUILT_IN_STRNCMP_EQ.
4223 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
4224 BUILT_IN_STRNCMP_EQ.
4225 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
4226 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4227 (gimple_fold_builtin): Likewise.
4228 * tree-ssa-strlen.c (compute_string_length): New function.
4229 (determine_min_obsize): New function.
4230 (handle_builtin_string_cmp): New function to handle calls to
4231 string compare functions.
4232 (strlen_optimize_stmt): Add handling to builtin string compare
4234 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
4235 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
4236 * tree.c (build_common_builtin_nodes): Add new defines of
4237 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
4239 2018-05-31 Jakub Jelinek <jakub@redhat.com>
4242 * bb-reorder.c (pass_partition_blocks::gate): Return false for
4243 functions with naked attribute.
4245 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
4247 * config/i386/sse.md (avx_vec_concat<mode>):
4248 Substitute concat_tg_mode mode attribute with xtg_mode.
4249 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
4250 (concat_tg_mode): Remove mode attribute.
4252 2018-05-31 Martin Sebor <msebor@redhat.com>
4255 * calls.c (alloc_max_size): Correct a logic error/typo.
4256 Treat excessive arguments as infinite. Warn for invalid arguments.
4257 * doc/invoke.texi (-Walloc-size-larger-than): Update.
4259 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
4262 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
4263 and movx for Haswell.
4265 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
4266 Cesar Philippidis <cesar@codesourcery.com>
4269 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
4270 when emitting error on private/firstprivate reductions.
4271 * omp-low.c (lower_omp_target): Avoid reference-type processing
4272 on pointers for firstprivate clause.
4274 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4276 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
4279 * config/aarch64/aarch64-simd.md
4280 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
4281 (aarch64_ld1_x3_<mode>): Likewise
4282 (aarch64_st1x2<VALLDIF:mode>): Likewise
4283 (aarch64_st1_x2_<mode>): Likewise
4284 (aarch64_st1x3<VALLDIF:mode>): Likewise
4285 (aarch64_st1_x3_<mode>): Likewise
4286 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
4287 (vld1_s8_x3): Likewise.
4288 (vld1_u16_x3): Likewise.
4289 (vld1_s16_x3): Likewise.
4290 (vld1_u32_x3): Likewise.
4291 (vld1_s32_x3): Likewise.
4292 (vld1_u64_x3): Likewise.
4293 (vld1_s64_x3): Likewise.
4294 (vld1_f16_x3): Likewise.
4295 (vld1_f32_x3): Likewise.
4296 (vld1_f64_x3): Likewise.
4297 (vld1_p8_x3): Likewise.
4298 (vld1_p16_x3): Likewise.
4299 (vld1_p64_x3): Likewise.
4300 (vld1q_u8_x3): Likewise.
4301 (vld1q_s8_x3): Likewise.
4302 (vld1q_u16_x3): Likewise.
4303 (vld1q_s16_x3): Likewise.
4304 (vld1q_u32_x3): Likewise.
4305 (vld1q_s32_x3): Likewise.
4306 (vld1q_u64_x3): Likewise.
4307 (vld1q_s64_x3): Likewise.
4308 (vld1q_f16_x3): Likewise.
4309 (vld1q_f32_x3): Likewise.
4310 (vld1q_f64_x3): Likewise.
4311 (vld1q_p8_x3): Likewise.
4312 (vld1q_p16_x3): Likewise.
4313 (vld1q_p64_x3): Likewise.
4314 (vst1_s64_x2): Likewise.
4315 (vst1_u64_x2): Likewise.
4316 (vst1_f64_x2): Likewise.
4317 (vst1_s8_x2): Likewise.
4318 (vst1_p8_x2): Likewise.
4319 (vst1_s16_x2): Likewise.
4320 (vst1_p16_x2): Likewise.
4321 (vst1_s32_x2): Likewise.
4322 (vst1_u8_x2): Likewise.
4323 (vst1_u16_x2): Likewise.
4324 (vst1_u32_x2): Likewise.
4325 (vst1_f16_x2): Likewise.
4326 (vst1_f32_x2): Likewise.
4327 (vst1_p64_x2): Likewise.
4328 (vst1q_s8_x2): Likewise.
4329 (vst1q_p8_x2): Likewise.
4330 (vst1q_s16_x2): Likewise.
4331 (vst1q_p16_x2): Likewise.
4332 (vst1q_s32_x2): Likewise.
4333 (vst1q_s64_x2): Likewise.
4334 (vst1q_u8_x2): Likewise.
4335 (vst1q_u16_x2): Likewise.
4336 (vst1q_u32_x2): Likewise.
4337 (vst1q_u64_x2): Likewise.
4338 (vst1q_f16_x2): Likewise.
4339 (vst1q_f32_x2): Likewise.
4340 (vst1q_f64_x2): Likewise.
4341 (vst1q_p64_x2): Likewise.
4342 (vst1_s64_x3): Likewise.
4343 (vst1_u64_x3): Likewise.
4344 (vst1_f64_x3): Likewise.
4345 (vst1_s8_x3): Likewise.
4346 (vst1_p8_x3): Likewise.
4347 (vst1_s16_x3): Likewise.
4348 (vst1_p16_x3): Likewise.
4349 (vst1_s32_x3): Likewise.
4350 (vst1_u8_x3): Likewise.
4351 (vst1_u16_x3): Likewise.
4352 (vst1_u32_x3): Likewise.
4353 (vst1_f16_x3): Likewise.
4354 (vst1_f32_x3): Likewise.
4355 (vst1_p64_x3): Likewise.
4356 (vst1q_s8_x3): Likewise.
4357 (vst1q_p8_x3): Likewise.
4358 (vst1q_s16_x3): Likewise.
4359 (vst1q_p16_x3): Likewise.
4360 (vst1q_s32_x3): Likewise.
4361 (vst1q_s64_x3): Likewise.
4362 (vst1q_u8_x3): Likewise.
4363 (vst1q_u16_x3): Likewise.
4364 (vst1q_u32_x3): Likewise.
4365 (vst1q_u64_x3): Likewise.
4366 (vst1q_f16_x3): Likewise.
4367 (vst1q_f32_x3): Likewise.
4368 (vst1q_f64_x3): Likewise.
4369 (vst1q_p64_x3): Likewise.
4371 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4373 * config/msp430/msp430.c (msp430_output_labelref): Prepend
4374 user_label_prefix to name.
4376 * tree-core.h: Update comment about the format of NAME string
4377 passed to handler in attribute_spec.
4379 * config/msp430/msp430.md: Remove erroneous subreg expression from
4380 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
4381 zero_extend{q,h}isi2.
4383 2018-05-30 Borislav Petkov <bp@suse.de>
4385 * doc/extend.texi: Document some architecture specific
4386 constraints and sort entries.
4388 2018-05-30 Martin Sebor <msebor@redhat.com>
4391 * builtins.c (expand_builtin_stpcpy_1): New function.
4392 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
4393 only if the former succeeds.
4395 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4397 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
4400 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4402 * doc/invoke.texi (-flinker-output): Document
4404 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4406 * passes.c (ipa_write_summaries): Only modify statements if body
4408 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
4409 incrementally linking.
4410 (ipa_passes): Likewise.
4411 * lto-cgraph.c (lto_output_node): When incrementally linking do not
4412 pass down resolution info.
4413 * common.opt (flag_incremental_link): Update info.
4414 * gcc.c (plugin specs): Turn flinker-output=* to
4415 -plugin-opt=-linker-output-known
4416 * toplev.c (compile_file): Also cut compilation when doing incremental
4418 * flag-types. (enum lto_partition_model): Add
4419 LTO_LINKER_OUTPUT_NOLTOREL.
4420 (invoke.texi): Add -flinker-output docs.
4421 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
4422 link same way as WPA; do not stream in dead initializers.
4424 * dwarf2out.c (dwarf2out_die_ref_for_decl,
4425 darf2out_register_external_decl): Support incremental link.
4427 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4429 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
4431 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
4433 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
4434 it down to simple_object_copy_lto_debug_sections.
4435 (run_gcc): Determine incremental LTO link time and configure
4436 lto1 into non-wpa mode, disable renaming of debug sections.
4438 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
4440 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4441 descriptions of various incorrectly documented functions.
4443 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
4446 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4447 address check not strict.
4449 2018-05-30 Richard Biener <rguenther@suse.de>
4451 PR tree-optimization/85964
4452 * tracer.c (better_p): Drop initialized count check, we only
4453 call the function with initialized counts now.
4454 (find_best_successor): Do find a best edge if one
4455 has uninitialized count.
4456 (find_best_predecessor): Likewise. Do BB frequency check only
4457 if count is initialized.
4459 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
4461 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4462 (aarch64_ldrstr_offset_compare): New.
4463 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
4464 load/store orderings.
4465 (aarch64_gen_adjusted_ldpstp): Likewise.
4467 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
4469 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4470 Check for subset of GENERAL_REGS and FP_REGS.
4471 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
4474 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
4476 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
4477 and wi::to_poly_offset. Add the current offset and then check
4478 whether the sum fits, rather than using an unchecked addition of
4479 a checked term. Check for a shwi rather than a uhwi.
4480 * expr.c (get_bit_range): Use tree_to_poly_uint64.
4481 (store_constructor): Use poly_int_tree_p.
4482 (expand_expr_real_1): Likewise.
4483 * function.c (assign_temp): Likewise.
4484 * fold-const.c (const_binop): Use poly_int_tree_p and
4486 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
4488 * ipa-icf-gimple.c (func_checker::compare_operand): Use
4489 to_poly_offset for MEM offsets.
4490 * ipa-icf.c (sem_variable::equals): Likewise.
4491 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
4492 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
4493 wi::to_poly_offset for BIT_FIELD_REF offsets.
4494 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
4496 * var-tracking.c (emit_note_insn_var_location): Use
4497 tree_to_poly_uint64.
4499 2018-05-29 Jim Wilson <jimw@sifive.com>
4501 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
4503 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
4506 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
4507 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
4509 (sse4_1_round<mode>2): Use nonimmediate_operand
4510 for operand 1 predicate.
4512 2018-05-29 Martin Sebor <msebor@redhat.com>
4513 Richard Biener <rguenther@suse.de>
4516 * calls.c (get_size_range): Call determine_value_range instead
4517 of get_value_range..
4518 * tree-vrp.h (determine_value_range): Declared new function.
4519 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
4521 2018-05-29 Richard Biener <rguenther@suse.de>
4523 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4524 sure to use non-pattern stmts for get_earlier_stmt arguments.
4525 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
4526 called on pattern stmts.
4527 (get_later_stmt): Likewise.
4529 2018-05-29 Martin Liska <mliska@suse.cz>
4531 PR gcov-profile/85759
4532 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
4535 2018-05-29 Jakub Jelinek <jakub@redhat.com>
4537 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
4539 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
4543 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4544 VEC_PACK_FLOAT_EXPR): New tree codes.
4545 * tree-pretty-print.c (op_code_prio): Handle
4546 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4547 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4548 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4549 * tree-inline.c (estimate_operator_cost): Likewise.
4550 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
4551 * fold-const.c (const_binop): Likewise.
4552 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
4553 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
4554 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4555 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
4556 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
4557 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
4558 * expr.c (expand_expr_real_2): Likewise.
4559 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
4560 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
4561 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
4563 * optabs.c (expand_widen_pattern_expr): For
4564 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
4565 sign from result type rather than operand's type.
4566 (expand_binop_directly): For vec_packu_float_optab and
4567 vec_packs_float_optab allow result type to be different from operand's
4569 * optabs-tree.c (optab_for_tree_code): Handle
4570 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4571 VEC_PACK_FLOAT_EXPR. Formatting fixes.
4572 * tree-vect-generic.c (expand_vector_operations_1): Handle
4573 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
4574 VEC_PACK_FLOAT_EXPR.
4575 * tree-vect-stmts.c (supportable_widening_operation): Handle
4577 (supportable_narrowing_operation): Handle FLOAT_EXPR.
4578 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
4579 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
4580 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
4581 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
4583 (vec_pack<floatprefix>_float_<mode>): New expander.
4584 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
4586 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
4587 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
4588 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
4589 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
4590 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
4592 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
4593 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
4594 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
4595 VEC_PACK_FLOAT_EXPR): Document.
4597 2018-05-29 Richard Biener <rguenther@suse.de>
4599 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
4601 (stmt_vec_info_vec): Make pointer.
4602 (init_stmt_vec_info_vec): Remove.
4603 (free_stmt_vec_info_vec): Likewise.
4604 (set_stmt_vec_info_vec): New function.
4605 (free_stmt_vec_infos): Likewise.
4606 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
4607 (set_vinfo_for_stmt): Likewise.
4608 (get_earlier_stmt): Likewise.
4609 (get_later_stmt): Likewise.
4610 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
4611 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
4612 (vec_info::~vec_info): Free stmt_vec_infos.
4613 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
4614 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
4615 (pass_slp_vectorize::execute): Likewise.
4616 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
4617 (free_stmt_vec_info_vec): Likewise.
4618 (set_stmt_vec_info_vec): New function.
4619 (free_stmt_vec_infos): Likewise.
4620 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
4621 the global stmt_vec_info_vec.
4622 * tree-parloops.c (gather_scalar_reductions): Use
4623 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
4626 2018-05-29 Richard Biener <rguenther@suse.de>
4628 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
4630 2018-05-29 Martin Liska <mliska@suse.cz>
4631 David Malcolm <dmalcolm@redhat.com>
4633 * vec.c (test_reverse): New.
4634 (vec_c_tests): Add new test.
4635 * vec.h (vl_ptr>::reverse): New function.
4637 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
4639 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
4641 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
4644 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
4646 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
4648 2018-05-28 Richard Biener <rguenther@suse.de>
4650 PR tree-optimization/85933
4651 * tree-vect-data-refs.c (vect_record_base_alignments): Only
4652 look at stmts marked as vectorizable.
4654 2018-05-28 Richard Biener <rguenther@suse.de>
4656 PR tree-optimization/85934
4657 * tree-vect-generic.c (expand_vector_operations_1): Hoist
4658 vector boolean check before scalar optimization.
4660 2018-05-28 Jakub Jelinek <jakub@redhat.com>
4662 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
4665 2018-05-28 Mark Wielaard <mark@klomp.org>
4667 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
4668 if it is an expression containing a minus sign.
4670 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
4672 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
4674 2018-05-27 Paul Koning <ni1d@arrl.net>
4676 * config/pdp11/pdp11.md (truncsihi2): Remove.
4678 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
4679 Chung-Ju Wu <jasonwucj@gmail.com>
4681 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
4683 (unaligned_store_dw): Ditto.
4684 * config/nds32/nds32-memory-manipulation.c
4685 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
4686 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
4687 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
4688 (emit_setmem_word_loop): Rename to ...
4689 (emit_setmem_doubleword_loop): ... this.
4690 (nds32_gen_dup_4_byte_to_word_value): New function.
4691 (nds32_gen_dup_8_byte_to_double_word_value): New function.
4692 (nds32_expand_setmem_loop): Refine implementation.
4693 (nds32_expand_setmem_loop_v3m): Ditto.
4694 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
4697 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4699 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
4701 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
4703 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
4704 (nds32_init_machine_status): Initialize machine->attr_naked_p and
4705 machine->attr_no_prologue_p.
4706 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
4707 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
4708 (nds32_expand_epilogue): Consider attr_naked_p.
4709 (nds32_expand_epilogue_v3pop): Likewise.
4710 (nds32_can_use_return_insn): Likewise.
4711 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
4712 attr_no_prologue_p fields.
4713 * config/nds32/nds32.opt (mret-in-naked-func): New option.
4715 2018-05-27 Jakub Jelinek <jakub@redhat.com>
4718 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
4720 * config/i386/sse.md
4721 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
4723 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
4725 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
4727 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
4729 (*<floatsuffix>floatv2div2sf2): Rename to ...
4730 (*float<floatunssuffix>v2div2sf2): ... this.
4731 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
4732 (float<floatunssuffix>v2div2sf2_mask): ... this.
4733 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
4734 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
4735 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
4737 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
4739 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4741 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
4743 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4745 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
4747 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
4748 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
4749 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
4750 gen_ufix_truncv8dfv8si2.
4751 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
4752 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
4753 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
4754 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
4755 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
4756 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
4757 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
4758 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
4760 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
4764 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
4766 2018-05-25 Jim Wilson <jimw@sifive.com>
4768 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
4769 * config/riscv/riscv.c (struct machine_function): Add
4770 interrupt_handler_p and attribute_checked_p fields.
4771 (riscv_attribute_table): Add interrupt.
4772 (riscv_interrupt_type_p): New.
4773 (riscv_save_reg_p): Save extra regs for interrupt handler.
4774 (riscv_use_save_libcall): Return false for interrupt handler.
4775 (riscv_first_stack_step): Add forward declaration.
4776 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
4777 for interrupt handler with large frame. Use it for saved reg list.
4778 (riscv_expand_prologue): Move flag_stack_usage_info support to
4779 eliminate duplication.
4780 (riscv_expand_epilogue): Generate mret for interrupt handler.
4781 (riscv_epilogue_uses): New.
4782 (riscv_can_use_return_insn): Return false for interrupt handler.
4783 (riscv_function_ok_for_sibcall): Likewise.
4784 (riscv_set_current_function): Add interrupt handler support.
4785 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
4786 * config/riscv/riscv.md (UNSPECV_MRET): New.
4788 (riscv_frflags, riscv_fsflags): Use tab after opcode.
4790 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
4792 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
4794 PR tree-optimization/85712
4795 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
4796 this candidate has already been replaced in-situ by a copy.
4798 2018-05-25 Jason Merrill <jason@redhat.com>
4800 PR c++/80485 - inline function non-zero address.
4801 * symtab.c (nonzero_address): Check DECL_COMDAT.
4803 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
4806 * config/alpha/alpha.md (ashlsi3): New insn pattern.
4807 (*ashlsi_se): Rename from *ashldi_se. Define as sign
4808 extension of SImode operation. Use const123_operand predicate.
4809 (*saddsi_1): Remove.
4810 (*saddl_se_1): Ditto.
4812 (*ssubl_se_1): Ditto.
4813 * config/alpha/predicates.md (const123_operand): New predicate.
4814 * config/alpha/constraints.md (P): Use IN_RANGE.
4816 2018-05-25 Richard Biener <rguenther@suse.de>
4818 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
4820 (ref_maybe_used_by_stmt_p): Likewise.
4821 (stmt_may_clobber_ref_p): Likewise.
4822 (stmt_may_clobber_ref_p_1): Likewise.
4823 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
4825 (ref_maybe_used_by_stmt_p): Likewise.
4826 (stmt_may_clobber_ref_p): Likewise.
4827 (stmt_may_clobber_ref_p_1): Likewise.
4828 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
4829 the alias oracle to disambiguate DRs with stmts DR analysis
4831 (vect_analyze_data_refs): Do not give up on not analyzable
4832 DRs for BB vectorization. Remove code truncating the dataref
4835 2018-05-25 Jakub Jelinek <jakub@redhat.com>
4838 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4839 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
4840 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
4842 2018-05-25 Richard Biener <rguenther@suse.de>
4844 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
4845 function, combining stmt data ref gathering and fatal analysis
4847 (vect_analyze_data_refs): Remove now redudnant code and simplify.
4848 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
4849 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
4850 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
4851 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
4853 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4855 PR tree-optimization/85720
4856 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
4857 SCC if all partitions are builtins.
4858 (version_loop_by_alias_check): New parameter. Generate cancelable
4859 runtime alias check if all partitions are builtins.
4860 (distribute_loop): Update call to above function.
4862 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4864 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
4865 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
4866 (parm_default_def_partition_arg): Ditto.
4867 (set_parm_default_def_partition): Ditto.
4868 (get_parm_default_def_partitions): Ditto and make it static.
4869 (get_undefined_value_partitions): Ditto and make it static.
4870 (remove_ssa_form): Refactor call to init_var_map here.
4871 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
4872 computation for loop region.
4873 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
4874 (register_default_def): Delete.
4875 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
4876 (parm_default_def_partition_arg): Ditto.
4877 (set_parm_default_def_partition): Ditto.
4878 (get_parm_default_def_partitions): Ditto and make it static.
4879 (get_undefined_value_partitions): Ditto and make it static.
4880 (coalesce_with_default, coalesce_with_default): Update comment.
4881 (create_coalesce_list_for_region): New func factored out from
4882 create_outofssa_var_map.
4883 (populate_coalesce_list_for_outofssa): New func factored out from
4884 create_outofssa_var_map and coalesce_ssa_name.
4885 (create_outofssa_var_map): Delete.
4886 (coalesce_ssa_name): Refactor to support live range computation.
4887 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
4888 (get_parm_default_def_partitions): Delete.
4889 (get_undefined_value_partitions): Ditto.
4890 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
4891 computation for loop region.
4892 (new_tree_live_info, loe_visit_block): Ditto.
4893 (live_worklist, set_var_live_on_entry): Ditto.
4894 (calculate_live_on_exit, verify_live_on_entry): Ditto.
4895 * tree-ssa-live.h (struct _var_map): New fields.
4896 (init_var_map): Change decl.
4897 (region_contains_p): New.
4899 2018-05-25 Bin Cheng <bin.cheng@arm.com>
4901 * tree-ssa-live.h (live_merge_and_clear): Delete.
4903 2018-05-25 Richard Biener <rguenther@suse.de>
4906 * tree-dump.c (dequeue_and_dump): Remove access to removed
4907 operand 2 of a SWITCH_EXPR.
4909 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4911 * doc/sourcebuild.texi (vect_double_cond_arith): Include
4912 multiplication and division.
4913 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
4914 (cond_udiv@var{m}, cond_umod@var{m}): Document.
4915 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
4916 (cond_udiv_optab, cond_umod_optab): New optabs.
4917 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
4918 (IFN_COND_RDIV): New internal functions.
4919 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
4920 TRUNC_MOD_EXPR and RDIV_EXPR.
4921 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
4922 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
4924 (SVE_INT_BINARY): Include mult.
4925 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
4926 (optab, sve_int_op): Handle mult.
4927 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
4929 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
4930 for SVE_INT_BINARY_SD.
4932 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4934 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
4935 (optab, sve_int_op): Handle div and udiv.
4936 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
4937 for SVE_INT_BINARY_SD.
4938 (*<optab><mode>3): New insn for the same.
4940 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4942 * tree-vect-patterns.c: Include predict.h.
4943 (vect_recog_divmod_pattern): Restrict check for division support
4944 to when optimizing for size.
4946 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
4948 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
4949 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
4950 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
4951 (gimple_match_op::set_op): Likewise.
4952 (gimple_resimplify4): Declare.
4953 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
4954 (expr::gen_transform): Likewise.
4955 (decision_tree::gen): Generate a simplification routine for 4 operands.
4956 * gimple-match-head.c (gimple_simplify): Add an overload for
4957 4 operands. In the top-level function, handle up to 4 call
4958 arguments and call gimple_resimplify4.
4959 (gimple_resimplify4): New function.
4960 (build_call_internal): Pass a fourth operand.
4961 (maybe_push_to_seq): Likewise.
4962 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
4963 Fold VEC_COND_EXPRs of an operation and a default value into
4964 an IFN_COND_* function if possible.
4965 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
4967 (SVE_COND_FP_BINARY): Include them.
4968 (optab, sve_fp_op): Handle them.
4969 (SVE_INT_BINARY_REV): New code iterator.
4970 (SVE_COND_FP_BINARY_REV): New int iterator.
4971 (commutative): New int attribute.
4972 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
4974 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
4976 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
4977 (*cond_<optab><mode>): New patterns for reversed operands.
4979 2018-05-25 Richard Biener <rguenther@suse.de>
4981 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
4982 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
4983 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
4984 (STMT_VINFO_GROUPED_ACCESS): Adjust.
4985 * tree-vect-data-refs.c (everywhere): Adjust users.
4986 * tree-vect-loop.c (everywhere): Likewise.
4987 * tree-vect-slp.c (everywhere): Likewise.
4988 * tree-vect-stmts.c (everywhere): Likewise.
4989 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4991 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4993 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
4995 (gcc_cv_as_section_exclude): ... this.
4996 Try Solaris as #exclude syntax.
4997 * configure: Regenerate.
4998 * config.in: Regenerate.
4999 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5001 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5002 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5004 * varasm.c (default_elf_asm_named_section): Don't check if
5005 HAVE_GAS_SECTION_EXCLUDE is defined.
5007 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5009 * doc/md.texi: Update the documentation of the cond_* optabs
5010 to mention the new final operand. Fix GET_MODE_NUNITS call.
5011 Describe the scalar case too.
5012 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5013 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5015 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5016 (get_conditional_internal_fn): Update comment.
5017 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5018 accumulator value as a final argument to conditional functions.
5019 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5020 a define_expand and add an "else" operand. Assert for now that
5021 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5022 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5023 (*cond_<optab><mode>): New patterns.
5024 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5025 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5026 (UNSPEC_COND_EOR): Delete.
5027 (optab): Remove associated mappings.
5028 (SVE_INT_BINARY): New code iterator.
5029 (sve_int_op): Remove int attribute and add "minus" to the code
5031 (SVE_COND_INT_OP): Delete.
5032 (SVE_COND_FP_OP): Rename to...
5033 (SVE_COND_FP_BINARY): ...this.
5035 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5037 * optabs.c (can_reuse_operands_p): New function.
5038 (maybe_legitimize_operands): Try to reuse the results for
5041 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5043 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5044 Add {q} suffix to insn mnemonic.
5046 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5048 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5049 (msp430_warn_func_return): New.
5051 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5053 * fold-const.c (tree_nonzero_bits): New function.
5054 * fold-const.h (tree_nonzero_bits): Likewise.
5055 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5056 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5058 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5062 * varasm.c (assemble_alias): Check ifunc_resolver only on
5065 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5068 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
5069 when memory input operand is handled.
5071 2018-05-24 Luis Machado <luis.machado@linaro.org>
5073 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
5075 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
5077 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5079 * match.pd: Delay FMA folds until after vectorization.
5081 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
5084 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5085 address check not strict.
5087 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
5089 * gimple-match.h (gimple_match_op): New class.
5090 (mprts_hook): Replace parameters with a gimple_match_op *.
5091 (maybe_build_generic_op): Likewise.
5092 (gimple_simplified_result_is_gimple_val): Replace parameters with
5093 a const gimple_match_op *.
5094 (gimple_simplify): Replace code_helper * and tree * parameters with
5095 a gimple_match_op * parameter.
5096 (gimple_resimplify1): Replace code_helper *, tree and tree *
5097 parameters with a gimple_match_op * parameter.
5098 (gimple_resimplify2): Likewise.
5099 (gimple_resimplify3): Likewise.
5100 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
5101 parameters with a gimple_match_op * parameter.
5102 * gimple-match-head.c (gimple_simplify): Change prototypes of
5103 auto-generated functions to take a gimple_match_op * instead of
5104 separate code_helper * and tree * parameters. Make the same
5105 change in the top-level overload and update calls to the
5106 gimple_resimplify routines. Update calls to the auto-generated
5107 functions and to maybe_push_res_to_seq in the publicly-facing
5108 operation-specific gimple_simplify overloads.
5109 (gimple_match_op::MAX_NUM_OPS): Define.
5110 (gimple_resimplify1): Replace rcode and ops with a single res_op
5111 parameter. Update call to gimple_simplify.
5112 (gimple_resimplify2): Likewise.
5113 (gimple_resimplify3): Likewise.
5114 (mprts_hook): Replace parameters with a gimple_match_op *.
5115 (maybe_build_generic_op): Likewise.
5116 (build_call_internal): Replace type, nargs and ops with
5117 a gimple_match_op *.
5118 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
5119 with a single gimple_match_op *. Update calls to mprts_hook,
5120 build_call_internal and gimple_simplified_result_is_gimple_val.
5121 Factor out code that is common to the tree_code and combined_fn cases.
5122 * genmatch.c (expr::gen_transform): Replace tem_code and
5123 tem_ops with a gimple_match_op called tem_op. Update calls
5124 to the gimple_resimplify functions and maybe_push_res_to_seq.
5125 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
5126 res_ops. Update call to the gimple_resimplify functions.
5127 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
5128 (decision_tree::gen): Make the functions take a gimple_match_op *
5129 called res_op instead of separate res_code and res_ops parameters.
5130 Update call accordingly.
5131 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
5132 and ops with a single res_op parameter. Update calls to
5133 maybe_build_generic_op and maybe_push_res_to_seq.
5134 (fold_stmt_1): Update calls to gimple_simplify and
5135 replace_stmt_with_simplification.
5136 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
5137 and gimple_simplified_result_is_gimple_val.
5138 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
5140 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
5141 with a gimple_match_op *.
5142 (vn_nary_build_or_lookup): Likewise. Update call to
5143 vn_nary_build_or_lookup_1.
5144 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
5145 gimple_match_op *. Update calls to the gimple_resimplify routines
5146 and to gimple_simplified_result_is_gimple_val.
5147 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
5148 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
5149 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
5150 (visit_nary_op): Likewise.
5151 (visit_reference_op_load): Likewise.
5153 2018-05-23 Luis Machado <luis.machado@linaro.org>
5155 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
5156 modifier for printing the step amount.
5158 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
5161 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
5164 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
5166 * doc/sourcebuild.texi (Endianness): New subsubsection.
5168 2018-05-23 Luis Machado <luis.machado@linaro.org>
5170 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5171 <prefetch_dynamic_strides>: New const bool field.
5172 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5173 prefetch_dynamic_strides.
5174 (exynosm1_prefetch_tune): Likewise.
5175 (thunderxt88_prefetch_tune): Likewise.
5176 (thunderx_prefetch_tune): Likewise.
5177 (thunderx2t99_prefetch_tune): Likewise.
5178 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
5180 (aarch64_override_options_internal): Update to set
5181 PARAM_PREFETCH_DYNAMIC_STRIDES.
5182 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5183 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5184 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5185 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5186 prefetch-dynamic-strides setting.
5188 2018-05-23 Luis Machado <luis.machado@linaro.org>
5190 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5191 <minimum_stride>: New const int field.
5192 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5193 minimum_stride field defaulting to -1.
5194 (exynosm1_prefetch_tune): Likewise.
5195 (thunderxt88_prefetch_tune): Likewise.
5196 (thunderx_prefetch_tune): Likewise.
5197 (thunderx2t99_prefetch_tune): Likewise.
5198 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
5199 <default_opt_level>: Set to 3.
5200 (aarch64_override_options_internal): Update to set
5201 PARAM_PREFETCH_MINIMUM_STRIDE.
5202 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5203 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5204 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5205 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5206 stride is constant and is below the minimum stride threshold.
5208 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5210 * config/arm/arm-cpus.in (mode26): Delete.
5211 (armv4): Delete mode26 reference.
5212 * config/arm/arm.c (arm_configure_build_target): Delete use of
5215 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5217 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
5219 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
5220 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
5222 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
5223 (floatunsdidf2): Ditto.
5225 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
5227 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
5228 (fixuns_trunc<mode>si2_avx512f): Ditto.
5229 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
5230 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
5231 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
5233 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5235 PR rtl-optimization/79985
5236 * df-scan.c (df_insn_refs_collect): Remove special case for
5237 global registers and asm statements.
5239 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
5241 * extend.texi (Global Register Variables): Rewrite the bullet list.
5242 Note that the register is available for allocation. Note that access
5243 via inline asm must use constraints. Add note about async-signal
5244 handlers. Remove paragraph about automagic register selection.
5246 2018-05-23 Richard Biener <rguenther@suse.de>
5248 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
5249 of fixed offset from memset VN.
5251 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
5253 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
5255 (alloc_cand_and_find_basis): Initialize first_interp field.
5256 (slsr_process_mul): Modify first_interp field.
5257 (slsr_process_add): Likewise.
5258 (slsr_process_cast): Modify first_interp field for each new
5260 (slsr_process_copy): Likewise.
5261 (dump_candidate): Dump first_interp field.
5262 (replace_mult_candidate): Process all interpretations, not just
5264 (replace_rhs_if_not_dup): Likewise.
5265 (replace_one_candidate): Likewise.
5267 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
5269 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5271 (aarch64_needs_frame_chain): New function.
5272 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
5274 2018-05-23 Sudakshina Das <sudi.das@arm.com>
5277 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
5278 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
5279 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
5280 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
5281 as true for strict-align.
5282 (aarch64_can_inline_p): Perform checks even when callee has no
5283 attributes to check for strict alignment.
5284 * doc/extend.texi (AArch64 Function Attributes): Document
5286 * doc/invoke.texi: (AArch64 Options): Likewise.
5288 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
5290 PR tree-optimization/85853
5291 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
5292 the handling of the root of the node to...
5293 (vect_slp_analyze_node_operations_1): ...this new function,
5294 and run the whole thing with the child nodes' def types
5295 set according to their SLP node's def type.
5297 2018-05-23 Richard Biener <rguenther@suse.de>
5300 * tree-data-ref.c (create_runtime_alias_checks): Defer
5301 and ignore overflow warnings.
5303 2018-05-23 Yury Gribov <tetra2005@gmail.com>
5305 PR tree-optimization/85822
5306 * tree-vrp.c (is_masked_range_test): Fix handling of negative
5309 2018-05-23 Richard Biener <rguenther@suse.de>
5311 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
5312 memset constants via native_interpret_expr.
5314 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
5317 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
5319 (cgraph_node::create_alias): Likewise.
5320 (cgraph_node::get_availability): Check ifunc_resolver instead
5321 of looking up ifunc attribute.
5322 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
5323 * varasm.c (do_assemble_alias): Likewise.
5324 (assemble_alias): Likewise.
5325 (default_binds_local_p_3): Likewise.
5326 * cgraph.h (cgraph_node): Add ifunc_resolver.
5327 (cgraph_node::only_called_directly_or_aliased_p): Return false
5329 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
5331 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
5332 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
5333 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
5334 instead of looking up ifunc attribute.
5336 2018-05-22 Luis Machado <luis.machado@linaro.org>
5338 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
5340 2018-05-22 Martin Sebor <msebor@redhat.com>
5343 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
5344 only when expasion succeeds.
5345 (expand_builtin_strcmp): Same.
5346 (expand_builtin_strncmp): Same.
5348 2018-05-22 Martin Sebor <msebor@redhat.com>
5350 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
5352 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5353 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5355 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5356 aarch64_mem_pair_operand with memory_operand and delete operand swapping
5358 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5359 Add check for legitimate_address.
5360 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
5361 (aarch64_swap_ldrstr_operands): New.
5362 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
5365 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
5366 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5368 * config/aarch64/aarch64.md: New patterns to generate stp
5370 (store_pair_sw, store_pair_dw): New patterns to generate stp for
5371 single words and double words.
5372 (load_pair_sw, load_pair_dw): Likewise.
5373 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
5375 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
5377 * config/aarch64/aarch64-ldpstp.md: Modify peephole
5378 for different mode ldpstp and add peephole for merged zero stores.
5380 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
5382 (aarch64_gen_store_pair): Rename calls to match new patterns.
5383 (aarch64_gen_load_pair): Rename calls to match new patterns.
5384 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
5385 (load_pair<DREG:mode><DREG2:mode>): ... This.
5386 (store_pair<mode>): Rename to...
5387 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
5388 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
5390 (V_INT_EQUIV): Handle SImode.
5391 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
5394 2018-05-22 Martin Sebor <msebor@redhat.com>
5397 * calls.c (maybe_warn_nonstring_arg): Use string length to set
5398 or ajust the presumed bound on an operation to avoid unnecessary
5401 2018-05-22 Martin Sebor <msebor@redhat.com>
5403 PR tree-optimization/85826
5404 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
5405 assuming that a DECL necesarily has a constant size.
5407 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
5410 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
5412 2018-05-22 Richard Biener <rguenther@suse.de>
5414 PR tree-optimization/85834
5415 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5416 non-constant and non-zero memset arguments.
5418 2018-05-22 Martin Liska <mliska@suse.cz>
5421 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
5423 2018-05-22 Richard Biener <rguenther@suse.de>
5425 PR tree-optimization/85863
5426 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
5427 comparisons when vectype is specified.
5428 (vectorizable_condition): Do not specify vectype for
5429 vect_is_simple_cond when SLP vectorizing.
5431 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
5434 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5435 define __ibm128 as long double.
5436 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
5437 as a distinct type when IEEE 128-bit support is enabled.
5438 (init_float128_ieee): Fix up conversions between IFmode and IEEE
5439 128-bit types to use the correct functions.
5440 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
5441 convert between 128-bit floating point types that have different
5442 modes but the same representation, instead of using gen_lowpart to
5444 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
5446 (IFKF_reg): New attributes to give the register constraints for
5448 (extend<mode>tf2_internal): New insns to mark an explicit
5449 conversion between 128-bit floating point types that have a
5450 different mode but share the same representation.
5452 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
5454 PR tree-optimization/85814
5455 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
5456 a null return from get_strinfo when unsharing the next
5457 strinfo in the chain.
5459 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5462 * varasm.c (weak_finish): Clean up weak_decls.
5464 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5466 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
5467 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
5468 UNSPEC_UADALP values.
5469 * config/aarch64/iterators.md (ABAL): New int iterator.
5472 (sur): Add mappings for the above.
5473 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
5475 (aarch64_<sur>abal<mode>_4): Likewise.
5476 (aarch64_<sur>adalp<mode>_3): Likewise.
5477 (<sur>sadv16qi): New define_expand.
5479 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
5481 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
5482 (*movdf_internal): Ditto.
5483 (*rcpsf2_sse): Ditto.
5484 (*rsqrtsf2_sse): Ditto.
5485 (*sqrt<mode>2_sse): Ditto.
5487 2018-05-21 Tamar Christina <tamar.christina@arm.com>
5489 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
5491 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
5492 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
5493 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
5494 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
5496 * config/aarch64/arm_neon.h: Likewise.
5497 * config/aarch64/iterators.md (VQ_I): New.
5499 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
5501 * config.gcc: Add arc/t-multilib-linux to tmake_file for
5503 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
5506 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5508 * config/nds32/constraints.md (S): New constraint.
5509 * config/nds32/nds32.md (call_internal): Use constraint S.
5510 (call_value_internal): Likewise.
5511 (sibcall_internal): Likewise.
5512 (sibcall_value_internal): Likewise.
5514 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5515 Chung-Ju Wu <jasonwucj@gmail.com>
5517 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
5520 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
5521 Chung-Ju Wu <jasonwucj@gmail.com>
5523 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
5524 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
5525 (nds32_rtx_costs_impl): Simplify.
5526 (nds32_address_cost_impl): Simplify.
5527 (nds32_init_rtx_costs): New function.
5528 (nds32_rtx_costs_speed_prefer): Likewise.
5529 (nds32_rtx_costs_size_prefer): Likewise.
5530 (nds32_address_cost_speed_prefer): Likewise.
5531 (nds32_address_cost_speed_fwprop): Likewise.
5532 (nds32_address_cost_size_prefer): Likewise.
5533 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
5534 * config/nds32/nds32.c (nds32_option_override): Use
5535 nds32_init_rtx_costs function.
5537 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
5539 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
5540 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
5541 (TARGET_PIPELINE_N8): Likewise.
5542 (TARGET_PIPELINE_N10): Likewise.
5543 (TARGET_PIPELINE_N13): Likewise.
5544 (TARGET_PIPELINE_GRAYWOLF): Likewise.
5546 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
5548 * config/nds32/nds32-fpu.md: Update copyright year.
5550 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5552 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
5554 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5556 * config/nds32/nds32.c
5557 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
5558 * config/nds32/nds32.opt (minline-asm-r15): New option.
5560 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5562 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
5564 * config/nds32/nds32.md (abssi2): New pattern.
5566 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
5568 * config/i386/i386.md (rex64namesuffix): New mode attribute.
5569 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
5570 Merge insn pattern from sse_cvtsi2ss<round_name> and
5571 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
5572 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
5573 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
5574 using SWI48 mode iterator.
5575 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
5576 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
5577 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
5578 pattern from sse_cvttss2si<round_saeonly_name>
5579 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
5580 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
5581 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
5582 using SWI48 mode iterator.
5583 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
5584 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
5585 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
5586 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
5587 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
5588 using SWI48 mode iterator.
5589 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
5590 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
5591 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
5592 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
5593 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
5594 SWI48 mode iterator.
5595 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
5596 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
5597 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
5598 pattern from sse_cvttsd2si<round_saeonly_name>
5599 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
5601 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5603 * config/nds32/nds32-md-auxiliary.c
5604 (nds32_valid_smw_lwm_base_p): Refine.
5605 (nds32_output_smw_single_word): Refine.
5606 (nds32_output_smw_double_word): New.
5607 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
5609 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5611 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
5612 (nds32_output_stack_pop): Refine.
5613 (nds32_expand_unaligned_load): Refine.
5614 (nds32_expand_unaligned_store): Refine.
5616 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5617 Chung-Ju Wu <jasonwucj@gmail.com>
5619 * config/nds32/constants.md: Add TP_REGNUM constant.
5620 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
5621 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
5623 * config/nds32/nds32-doubleword.md: Consider flag_pic.
5624 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
5625 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
5626 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
5627 and PIC code generation.
5628 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
5630 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
5632 * config/nds32/nds32.md: Support TLS and PIC.
5633 * config/nds32/nds32.c: Support TLS and PIC.
5634 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
5635 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
5638 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
5640 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
5641 mode with E_ prefix.
5643 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5644 Chung-Ju Wu <jasonwucj@gmail.com>
5646 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
5647 * config/nds32/nds32-md-auxiliary.c
5648 (symbolic_reference_mentioned_p): New.
5649 (nds32_legitimize_ict_address): New.
5650 (nds32_expand_ict_move): New.
5651 (nds32_indirect_call_referenced_p): New.
5652 (nds32_symbol_binds_local_p): Delete.
5653 (nds32_long_call_p): Modify.
5654 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
5655 * config/nds32/nds32-protos.h
5656 (symbolic_reference_mentioned_p): Declare.
5657 (nds32_legitimize_ict_address): Declare.
5658 (nds32_expand_ict_move): Declare.
5659 (nds32_indirect_call_referenced_p): Declare.
5660 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
5661 (nds32_relax_group): Use nds32_ict_const_p as condition.
5662 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
5663 (nds32_asm_file_start): Output ict_model directive in asm code.
5664 (nds32_legitimate_address_p): Consider indirect call.
5665 (nds32_print_operand): Consider indirect call.
5666 (nds32_print_operand_address): Consider indirect call.
5667 (nds32_insert_attributes): Handle "indirect_call" attribute.
5668 (TARGET_LEGITIMATE_ADDRESS_P): Define.
5669 (TARGET_LEGITIMATE_CONSTANT_P): Define.
5670 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
5671 (TARGET_DELEGITIMIZE_ADDRESS): Define.
5672 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
5673 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
5674 (TARGET_ICT_MODEL_SMALL): Define.
5675 (TARGET_ICT_MODEL_LARGE): Define.
5676 * config/nds32/nds32.md (movsi): Consider ict model.
5677 (call, call_value): Consider ict model.
5678 (sibcall, sibcall_value): Consider ict model.
5679 * config/nds32/nds32.opt (mict-model): New option.
5680 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
5683 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
5684 Monk Chiang <sh.chiang04@gmail.com>
5685 Jim Wilson <jimw@sifive.com>
5687 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
5688 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
5689 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
5690 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
5691 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
5692 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
5693 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
5694 compute save_libcall_adjustment properly.
5695 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
5696 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
5697 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
5698 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
5699 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
5700 (ABI_SPEC): Handle mabi=ilp32e.
5701 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
5702 (RVE): Add RVE mask.
5703 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
5704 <-march>: Add rv32e as an example.
5706 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
5709 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
5710 (intra_create_variable_infos): Handle C++ constructors.
5712 2018-05-18 Martin Liska <mliska@suse.cz>
5714 * passes.def: Remove a redundant pass.
5716 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
5719 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
5721 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5723 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
5725 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
5727 (armv2, armv2a, armv3, armv3m): Delete architectures.
5728 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
5729 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
5730 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
5732 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
5733 (*mulsidi3adddi): Likewise.
5734 (mulsidi3): Likewise.
5735 (*mulsidi3_nov6): Likewise.
5736 (umulsidi3): Likewise.
5737 (umulsidi3_nov6): Likewise.
5738 (umaddsidi4): Likewise.
5739 (*umulsidi3adddi): Likewise.
5740 (smulsi3_highpart): Likewise.
5741 (*smulsi3_highpart_nov6): Likewise.
5742 (umulsi3_highpart): Likewise.
5743 (*umulsi3_highpart_nov6): Likewise.
5744 * config/arm/arm.h (arm_arch3m): Delete.
5745 * config/arm/arm.c (arm_arch3m): Delete.
5746 (arm_option_override_internal): Update armv3-related comment.
5747 (arm_configure_build_target): Delete use of isa_bit_mode32.
5748 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
5749 (arm_rtx_costs_internal): Delete check of arm_arch3m.
5750 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
5752 (mulusa3): Likewise.
5753 * config/arm/arm-protos.h (arm_arch3m): Delete.
5754 * config/arm/arm-tables.opt: Regenerate.
5755 * config/arm/arm-tune.md: Likewise.
5756 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
5757 deleted architectures.
5759 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5761 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
5762 (armv5t, armv5te): New features.
5763 (ARMv5, ARMv5e): Delete fgroups.
5764 (ARMv5t, ARMv5te): Adjust for above changes.
5766 (armv5, armv5e): Delete arches.
5767 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
5769 (*call_reg_arm): Likewise.
5770 (*call_value_reg_armv5): Likewise.
5771 (*call_value_reg_arm): Likewise.
5772 (*call_symbol): Likewise.
5773 (*call_value_symbol): Likewise.
5774 (*sibcall_insn): Likewise.
5775 (*sibcall_value_insn): Likewise.
5777 (prefetch): Likewise.
5778 (define_split and define_peephole2 dependent on arm_arch5):
5780 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
5782 (TARGET_ARM_QBIT): Likewise.
5783 (TARGET_DSP_MULTIPLY): Likewise.
5784 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
5785 (arm_arch5, arm_arch5e): Delete.
5786 (arm_arch5t, arm_arch5te): Declare.
5787 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
5788 (arm_arch5t): Declare.
5789 (arm_option_reconfigure_globals): Update for the above.
5790 (arm_options_perform_arch_sanity_checks): Update comment, replace
5791 use of arm_arch5 with arm_arch5t.
5792 (use_return_insn): Likewise.
5793 (arm_emit_call_insn): Likewise.
5794 (output_return_instruction): Likewise.
5795 (arm_final_prescan_insn): Likewise.
5796 (arm_coproc_builtin_available): Likewise.
5797 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
5798 arm_arch5e with arm_arch5t and arm_arch5te.
5799 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
5800 (arm_arch5t, arm_arch5te): Declare.
5801 * config/arm/arm-tables.opt: Regenerate.
5802 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
5803 * config/arm/t-multilib: Likewise.
5804 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
5805 instead of arm_arch5.
5806 (*call_reg_thumb1): Likewise.
5807 (*call_value_reg_thumb1_v5): Likewise.
5808 (*call_value_reg_thumb1): Likewise.
5809 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
5811 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
5813 2018-05-18 Martin Liska <mliska@suse.cz>
5815 PR gcov-profile/84846
5816 * doc/gcov.texi: Document -t option of gcov tool.
5818 2018-05-18 Martin Liska <mliska@suse.cz>
5820 PR gcov-profile/84846
5821 * gcov.c (print_usage): Add new -t option.
5822 (process_args): Handle the option.
5823 (generate_results): Use stdout as output when requested by
5826 2018-05-18 Martin Liska <mliska@suse.cz>
5828 PR gcov-profile/84846
5829 * coverage.c (coverage_init): Write PWD to .gcno file.
5830 * doc/gcov.texi: Document how working directory is printed.
5831 * gcov-dump.c (dump_gcov_file): Print PWD.
5832 * gcov.c (output_intermediate_file): Likewise.
5833 (read_graph_file): Read PWD string.
5834 (output_lines): Print PWD.
5836 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5839 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
5840 for retval and return false if all args to phi are zero.
5842 2018-05-18 Richard Biener <rguenther@suse.de>
5844 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
5846 (evrp_dom_walker::before_dom_children): Call it.
5848 2018-05-18 Richard Biener <rguenther@suse.de>
5850 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
5851 results when processing array refs with variable index.
5853 2018-05-18 Toon Moene <toon@moene.org>
5855 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
5856 directly after that of -floop-interchange. Indicate that both
5857 options are enabled by default when specifying -O3.
5859 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5861 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
5862 iterator. Delete separate integer-mode vec_set<mode> expander.
5863 (aarch64_simd_vec_setv2di): Delete.
5864 (vec_setv2di): Delete.
5865 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
5866 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
5867 the "w, r" alternative.
5869 2018-05-18 Martin Liska <mliska@suse.cz>
5871 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
5872 * tree-pass.h (make_pass_lower_switch_O0): New function.
5873 * tree-switch-conversion.c (node_has_low_bound): Remove.
5874 (node_has_high_bound): Likewise.
5875 (node_is_bounded): Likewise.
5876 (class pass_lower_switch): Make it a template type and create
5878 (pass_lower_switch::execute): Add template argument.
5879 (make_pass_lower_switch): New function.
5880 (make_pass_lower_switch_O0): New function.
5881 (do_jump_if_equal): Remove.
5882 (emit_case_nodes): Simplify to just handle all 3 cases and leave
5883 all the hard work to tree optimization passes.
5885 2018-05-18 Martin Liska <mliska@suse.cz>
5887 * dbgcnt.c (limit_low): Renamed from limit.
5888 (limit_high): New variable.
5889 (dbg_cnt_is_enabled): Check for upper limit.
5890 (dbg_cnt): Adjust dumping.
5891 (dbg_cnt_set_limit_by_index): Add new argument for high
5893 (dbg_cnt_set_limit_by_name): Likewise.
5894 (dbg_cnt_process_single_pair): Parse new format.
5895 (dbg_cnt_process_opt): Use strtok.
5896 (dbg_cnt_list_all_counters): Remove 'value' and add
5898 * doc/invoke.texi: Document changes.
5900 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
5902 * doc/sourcebuild.texi (scalar_all_fma): Document.
5903 * tree.def (FMA_EXPR): Delete.
5904 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
5905 * internal-fn.c (ternary_direct): New macro.
5906 (expand_ternary_optab_fn): Likewise.
5907 (direct_ternary_optab_supported_p): Likewise.
5908 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
5909 * builtins.c (fold_builtin_fma): Delete.
5910 (fold_builtin_3): Don't call it.
5911 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
5912 * expr.c (expand_expr_real_2): Likewise.
5913 * fold-const.c (operand_equal_p): Likewise.
5914 (fold_ternary_loc): Likewise.
5915 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
5916 * gimple.c (DEFTREECODE): Likewise.
5917 * gimplify.c (gimplify_expr): Likewise.
5918 * optabs-tree.c (optab_for_tree_code): Likewise.
5919 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
5920 * tree-eh.c (operation_could_trap_p): Likewise.
5921 (stmt_could_throw_1_p): Likewise.
5922 * tree-inline.c (estimate_operator_cost): Likewise.
5923 * tree-pretty-print.c (dump_generic_node): Likewise.
5924 (op_code_prio): Likewise.
5925 * tree-ssa-loop-im.c (stmt_cost): Likewise.
5926 * tree-ssa-operands.c (get_expr_operands): Likewise.
5927 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
5928 * fold-const-call.h (fold_fma): Delete.
5929 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
5930 CFN_FNMA and CFN_FNMS.
5932 * genmatch.c (combined_fn): New enum.
5933 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
5934 (commutative_op): New function.
5935 (commutate): Use it. Handle more than 2 operands.
5936 (dt_operand::gen_gimple_expr): Use commutative_op.
5937 (parser::parse_expr): Allow :c to be used with non-binary
5938 operators if the commutative operand is known.
5939 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
5940 CFN_FMS, CFN_FNMA and CFN_FNMS.
5941 (backprop::process_assign_use): Remove FMA_EXPR handling.
5942 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
5943 (gen_hsa_fma): New function.
5944 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
5945 IFN_FNMA and IFN_FNMS.
5946 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
5947 * gimple-fold.h (follow_all_ssa_edges): Declare.
5948 * gimple-fold.c (follow_all_ssa_edges): New function.
5949 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
5950 gimple_build interface and use follow_all_ssa_edges to fold the result.
5951 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
5952 instead of checking for optabs directly.
5953 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
5954 rather than FMA_EXPRs.
5955 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
5956 call to IFN_FMA instead of an FMA_EXPR.
5958 2018-05-17 Jim Wilson <jimw@sifive.com>
5960 * expr.c (do_tablejump): When converting index to Pmode, if we have a
5961 sign extended promoted subreg, and the range does not have the sign bit
5962 set, then do a sign extend.
5964 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
5965 test, check for sign extended subreg and/or constant operands, and
5966 do a sign extend in that case.
5968 2018-05-17 Steve Ellcey <sellcey@cavium.com>
5970 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
5971 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
5973 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
5974 Change logics_shift_reg to logics_shift_imm.
5975 (thunderx2t99_fp_loadpair_basic): Delete.
5976 (thunderx2t99_fp_storepair_basic): Delete.
5977 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
5978 (thunderx2t99_asimd_polynomial): Delete.
5979 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
5980 and neon_fp_mul_d_scalar_q.
5981 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
5982 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
5983 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
5984 (thunderx2t99_asimd_lut): Add missing tbl types.
5985 (thunderx2t99_asimd_ext): Delete.
5986 (thunderx2t99_asimd_load1_1_mult): Delete.
5987 (thunderx2t99_asimd_load1_2_mult): Delete.
5988 (thunderx2t99_asimd_load1_ldp): New.
5989 (thunderx2t99_asimd_load1): New.
5990 (thunderx2t99_asimd_load2): Add missing *load2* types.
5991 (thunderx2t99_asimd_load3): New.
5992 (thunderx2t99_asimd_load4): New.
5993 (thunderx2t99_asimd_store1_1_mult): Delete.
5994 (thunderx2t99_asimd_store1_2_mult): Delete.
5995 (thunderx2t99_asimd_store2_mult): Delete.
5996 (thunderx2t99_asimd_store2_onelane): Delete.
5997 (thunderx2t99_asimd_store_stp): New.
5998 (thunderx2t99_asimd_store1): New.
5999 (thunderx2t99_asimd_store2): New.
6000 (thunderx2t99_asimd_store3): New.
6001 (thunderx2t99_asimd_store4): New.
6003 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6005 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6006 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6008 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6009 Segher Boessenkool <segher@kernel.crashing.org>
6012 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6015 2018-05-17 Richard Biener <rguenther@suse.de>
6017 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6018 for pruning loop and prune defs feeding only already visited PHIs.
6020 2018-05-17 Richard Biener <rguenther@suse.de>
6022 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6024 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6025 Richard Biener <rguenther@suse.de>
6027 PR tree-optimization/85793
6028 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6029 for VMAT_ELEMENTWISE.
6031 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6033 * internal-fn.h (lookup_internal_fn): Declare
6034 * internal-fn.c (lookup_internal_fn): New function.
6035 * gimple.c (gimple_build_call_from_tree): Handle calls to
6037 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6038 internal function names.
6039 * tree-pretty-print.c (dump_generic_node): Likewise.
6040 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6042 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6044 * gimple-fold.h (gimple_build): Make the function forms take
6045 combined_fn rather than built_in_function.
6046 (gimple_simplify): Likewise.
6047 * gimple-match-head.c (gimple_simplify): Likewise.
6048 * gimple-fold.c (gimple_build): Likewise.
6049 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6050 rather than gimple_build_call_internal.
6051 (get_initial_defs_for_reduction): Likewise.
6052 (vect_create_epilog_for_reduction): Likewise.
6053 (vectorizable_live_operation): Likewise.
6055 2018-05-17 Martin Liska <mliska@suse.cz>
6057 * gimple-ssa-sprintf.c (format_directive): Do not use
6058 space in between 'G_' and '('.
6060 2018-05-17 Jakub Jelinek <jakub@redhat.com>
6063 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
6064 even if the mask is not all ones.
6067 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6069 (ix86_gimple_fold_builtin): Likewise.
6072 * config/i386/i386.c: Include tree-vector-builder.h.
6073 (ix86_vector_shift_count): New function.
6074 (ix86_fold_builtin): Fold shift builtins by scalar count.
6075 (ix86_gimple_fold_builtin): Likewise.
6077 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
6078 _mm512_setzero): New intrinsics.
6080 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
6081 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6083 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
6084 code generation for cases where splatting a value is not useful.
6085 * simplify-rtx.c (simplify_ternary_operation): Simplify
6086 vec_merge across a vec_duplicate and a paradoxical subreg forming
6087 a vector mode to a vec_concat.
6089 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
6091 * config.gcc: Support "goldmont-plus".
6092 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6094 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6095 PROCESSOR_GOLDMONT_PLUS.
6096 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
6097 (processor_target_table): Add "goldmont-plus".
6098 (PTA_GOLDMONT_PLUS): Define.
6099 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
6100 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
6101 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
6102 (fold_builtin_cpu): Add "goldmont-plus".
6103 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
6104 (ix86_option_override_internal): Add "goldmont-plus".
6105 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
6106 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
6107 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
6108 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
6110 2018-05-17 Richard Biener <rguenther@suse.de>
6112 PR tree-optimization/85757
6113 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
6114 remove defs that only feed that PHI from further processing.
6116 2018-05-16 Jim Wilson <jimw@sifive.com>
6118 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
6120 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
6122 2018-05-16 Mark Wielaard <mark@klomp.org>
6124 * dwarf2out.c (count_index_strings): New function.
6125 (output_indirect_strings): Call count_index_strings and generate
6126 header for dwarf_version >= 5.
6128 2018-05-16 Mark Wielaard <mark@klomp.org>
6130 * dwarf2out.c (dwarf_FORM): New function.
6131 (set_indirect_string): Use dwarf_FORM.
6132 (reset_indirect_string): Likewise.
6133 (size_of_die): Likewise.
6134 (value_format): Likewise.
6135 (output_die): Likewise.
6136 (add_skeleton_AT_string): Likewise.
6137 (output_macinfo_op): Likewise.
6138 (index_string): Likewise.
6139 (output_index_string_offset): Likewise.
6140 (output_index_string): Likewise.
6141 (count_index_strings): Likewise.
6143 2018-05-16 Carl Love <cel@us.ibm.com>
6145 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6146 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
6148 2018-05-16 Martin Jambor <mjambor@suse.cz>
6150 * ipa-prop.c (ipa_free_all_edge_args): Remove.
6151 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
6153 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
6155 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
6156 (fnma<mode>4): Likewise.
6157 (fms<mode>4): Likewise.
6158 (fnms<mode>4): Likewise.
6159 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
6160 (aarch64_fnma<mode>4): Likewise.
6161 (aarch64_fms<mode>4): Likewise.
6162 (aarch64_fnms<mode>4): Likewise.
6163 (aarch64_fnmadd<mode>4): Likewise.
6165 2018-05-16 Jason Merrill <jason@redhat.com>
6167 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
6169 2018-05-16 Richard Biener <rguenther@suse.de>
6171 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
6172 (dump_stmt_cost): Declare.
6173 (add_stmt_cost): Dump cost we add.
6174 (add_stmt_costs): New function.
6175 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
6177 (vect_analyze_stmt): Adjust prototype.
6178 (vectorizable_condition): Likewise.
6179 (vectorizable_live_operation): Likewise.
6180 (vectorizable_reduction): Likewise.
6181 (vectorizable_induction): Likewise.
6182 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
6183 cost vector to pass to vectorizable_ and record afterwards.
6184 (vect_model_reduction_cost): Take cost vector argument and adjust.
6185 (vect_model_induction_cost): Likewise.
6186 (vectorizable_reduction): Likewise.
6187 (vectorizable_induction): Likewise.
6188 (vectorizable_live_operation): Likewise.
6189 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
6190 SLP_TREE_NUMBER_OF_VEC_STMTS.
6191 (vect_analyze_slp_cost_1): Remove.
6192 (vect_analyze_slp_cost): Likewise.
6193 (vect_slp_analyze_node_operations): Take visited args and
6194 a target cost vector. Avoid processing already visited stmt sets.
6195 (vect_slp_analyze_operations): Use a local cost vector to gather
6196 costs and register those of non-discarded instances.
6197 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
6198 (vect_schedule_slp_instance): Remove copying of
6199 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
6201 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
6202 adding cost. Record cost entry location.
6203 (vect_prologue_cost_for_slp_op): Function to compute cost of
6204 a constant or invariant generated for SLP vect in the prologue,
6205 split out from vect_analyze_slp_cost_1.
6206 (vect_model_simple_cost): Make static. Adjust for SLP costing.
6207 (vect_model_promotion_demotion_cost): Likewise.
6208 (vect_model_store_cost): Likewise, make static.
6209 (vect_model_load_cost): Likewise.
6210 (vectorizable_bswap): Add cost vector arg and adjust.
6211 (vectorizable_call): Likewise.
6212 (vectorizable_simd_clone_call): Likewise.
6213 (vectorizable_conversion): Likewise.
6214 (vectorizable_assignment): Likewise.
6215 (vectorizable_shift): Likewise.
6216 (vectorizable_operation): Likewise.
6217 (vectorizable_store): Likewise.
6218 (vectorizable_load): Likewise.
6219 (vectorizable_condition): Likewise.
6220 (vectorizable_comparison): Likewise.
6221 (can_vectorize_live_stmts): Likewise.
6222 (vect_analyze_stmt): Likewise.
6223 (vect_transform_stmt): Adjust calls to vectorizable_*.
6224 * tree-vectorizer.c: Include gimple-pretty-print.h.
6225 (dump_stmt_cost): New function.
6227 2018-05-16 Richard Biener <rguenther@suse.de>
6229 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
6230 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
6231 * tree-ssa-dse.c: Include tree-ssa-loop.h.
6232 (check_name): New callback.
6233 (dse_classify_store): Track cycles via a visited bitmap of PHI
6234 defs and simplify handling of in-loop and across loop dead stores
6235 and properly fail for loop-variant refs. Handle byte-tracking with
6236 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
6239 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
6241 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
6242 (vect_get_mask_type_for_stmt): Likewise.
6243 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
6245 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
6246 to determine the statement's vector type and the vector type that
6247 should be used for calculating nunits. Deal with cases in which
6248 the type has to be deferred.
6249 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
6250 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
6251 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
6252 (vect_determine_vf_for_stmt): New functions, split out from...
6253 (vect_determine_vectorization_factor): ...here.
6254 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
6255 (vect_get_mask_type_for_stmt): New functions, split out from
6256 vect_determine_vectorization_factor.
6258 2018-05-16 Richard Biener <rguenther@suse.de>
6260 * tree-cfg.c (verify_gimple_assign_ternary): Properly
6261 verify the [VEC_]COND_EXPR embedded comparison.
6263 2018-05-15 Martin Sebor <msebor@redhat.com>
6265 PR tree-optimization/85753
6266 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
6267 RECORD_TYPE in addition to ARRAY_TYPE.
6269 2018-05-15 Martin Sebor <msebor@redhat.com>
6272 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
6274 2018-05-15 Richard Biener <rguenther@suse.de>
6276 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
6277 add by_clobber_p one. Change algorithm to collect all defs
6278 representing uses we need to walk and try reducing them to
6279 a single one before failing.
6280 (dse_dom_walker::dse_optimize_stmt): Adjust.
6282 2018-05-13 Mark Wielaard <mark@klomp.org>
6284 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
6285 (size_of_loc_descr): Likewise.
6286 (output_loc_operands): Likewise.
6287 (output_loc_operands_raw): Likewise.
6288 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
6289 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
6290 (hash_loc_operands): Likewise.
6291 (compare_loc_operands): Likewise.
6293 2018-05-14 Mark Wielaard <mark@klomp.org>
6295 * dwarf2out.c (count_index_addrs): New function.
6296 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
6298 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6300 PR tree-optimization/83648
6301 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
6302 return value as malloc candidate.
6304 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6307 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
6308 param as true in call to suggest_attribute.
6310 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
6312 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
6313 -mreadonly-in-sdata.
6315 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6317 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
6319 (aarch64_crypto_aesd_fused): Likewise.
6321 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
6323 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
6324 (movsi_aarch64): Likewise.
6325 (load_pairsi): Likewise.
6326 (load_pairdi): Likewise.
6327 (store_pairsi): Likewise.
6328 (store_pairdi): Likewise.
6329 (load_pairsf): Likewise.
6330 (load_pairdf): Likewise.
6331 (store_pairsf): Likewise.
6332 (store_pairdf): Likewise.
6333 (zero_extend): Likewise.
6334 (trunc): Swap alternatives.
6335 (fcvt_target): Add '?' to prefer w over r.
6337 2018-05-14 Jakub Jelinek <jakub@redhat.com>
6340 * config/i386/i386.md: Disallow non-commutative arithmetics in
6341 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
6342 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
6343 in the peephole2 before it.
6345 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
6347 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
6348 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
6349 (ix86_handle_option): Handle -mcldemote.
6350 * config.gcc: New header.
6351 * config/i386/cldemoteintrin.h: New file.
6352 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
6353 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
6355 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6356 OPTION_MASK_ISA_CLDEMOTE.
6357 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
6358 (ix86_valid_target_attribute_inner_p): Ditto.
6359 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
6360 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
6361 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
6362 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
6363 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
6365 * config/i386/i386.opt: Add -mcldemote.
6366 * config/i386/x86intrin.h: New header.
6367 * doc/invoke.texi: Add -mcldemote.
6369 2018-05-14 Richard Biener <rguenther@suse.de>
6371 * doc/match-and-simplify.texi: Adjust :s documentation.
6373 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
6375 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
6376 intended memcpy size.
6377 (REORDER_45): Likewise.
6379 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
6381 * sort.cc: New file.
6382 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
6383 * vec.c (qsort_chk): Use gcc_qsort.
6384 * Makefile.in (OBJS-libcommon): Add sort.o.
6385 (build/sort.o): New target. Use it...
6386 (BUILD_RTL): ... here, and...
6387 (build/gencfn-macros): ... here, and...
6388 (build/genmatch): ... here.
6390 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6391 Chung-Ju Wu <jasonwucj@gmail.com>
6393 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
6394 * config/nds32/nds32-graywolf.md: New file.
6395 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
6396 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
6398 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
6399 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
6400 * config/nds32/nds32.md (pipeline_model): Add graywolf.
6401 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
6402 * config/nds32/pipelines.md: Include n15 settings.
6404 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6405 Chung-Ju Wu <jasonwucj@gmail.com>
6407 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
6408 * config/nds32/nds32-n13.md: New file.
6409 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
6410 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
6412 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
6413 * config/nds32/nds32.md (pipeline_model): Add n13.
6414 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
6415 * config/nds32/pipelines.md: Include n13 settings.
6417 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
6418 Chung-Ju Wu <jasonwucj@gmail.com>
6420 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
6421 * config/nds32/nds32-n10.md: New file.
6422 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
6423 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
6425 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
6426 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
6427 * config/nds32/nds32.md (pipeline_model): Add n10.
6428 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
6429 * config/nds32/pipelines.md: Include n10 settings.
6431 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
6432 Kito Cheng <kito.cheng@gmail.com>
6433 Chung-Ju Wu <jasonwucj@gmail.com>
6435 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
6436 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6437 Add enum values for DSP extension instructions.
6438 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
6440 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
6441 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
6443 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
6444 * config/nds32/nds32-dspext.md: New file for DSP implementation.
6445 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
6446 * config/nds32/nds32-intrinsic.md: Likewise.
6447 * config/nds32/nds32_intrinsic.h: Likewise.
6448 * config/nds32/nds32-md-auxiliary.c: Likewise.
6449 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
6450 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
6451 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
6452 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
6453 * config/nds32/nds32-protos.h: New declarations for DSP extension.
6454 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
6455 TYPE_DMAC in switch statement.
6456 * config/nds32/nds32.c: New checking and implementation for DSP
6457 extension instructions.
6458 * config/nds32/nds32.h: Likewise.
6459 * config/nds32/nds32.md: Likewise.
6460 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
6461 * config/nds32/predicates.md: Implement new predicates for DSP
6464 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
6466 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
6467 Reformat alternatives and attributes so it is easier to identify
6468 which constraints/attributes go with which instruction.
6469 (mov<mode>_hardfloat32, FMOVE64): Likewise.
6470 (mov<mode>_softfloat32, FMOVE64): Likewise.
6471 (mov<mode>_hardfloat64, FMOVE64): Likewise.
6472 (mov<mode>_softfloat64, FMOVE64): Likewise.
6474 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
6476 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6478 (Basic PowerPC Built-in Functions): The new name of the
6479 subsection previously known as "PowerPC Built-in Functions".
6480 (Basic PowerPC Built-in Functions Available on all Configurations):
6482 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
6483 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
6484 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
6485 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
6487 2018-05-11 Martin Jambor <mjambor@suse.cz>
6490 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
6493 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6496 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
6498 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
6500 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
6501 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
6502 (ix86_handle_option): Handle -mwaitpkg.
6503 * config.gcc: New header.
6504 * config/i386/cpuid.h (bit_WAITPKG): New bit.
6505 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
6506 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
6508 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6509 OPTION_MASK_ISA_WAITPKG.
6510 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
6511 (ix86_option_override_internal): Add PTA_WAITPKG.
6512 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
6513 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
6514 IX86_BUILTIN_TPAUSE.
6515 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
6516 __builtin_ia32_umwait and __builtin_ia32_tpause.
6517 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
6518 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
6519 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
6520 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
6521 UNSPECV_TPAUSE): New.
6522 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
6523 * config/i386/i386.opt: Add -mwaitpkg.
6524 * config/i386/waitpkgintrin.h: New file.
6525 * config/i386/x86intrin.h: New header.
6526 * doc/invoke.texi: Add -mwaitpkg.
6528 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
6531 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
6533 (cortex-m0): Use armv6s-m isa.
6534 (cortex-m0plus): Likewise.
6535 (cortex-m1): Likewise.
6536 (cortex-m0.small-multiply): Likewise.
6537 (cortex-m0plus.small-multiply): Likewise.
6538 (cortex-m1.small-multiply): Likewise.
6540 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
6541 Jakub Jelinek <jakub@redhat.com>
6543 PR tree-optimization/85692
6544 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
6545 source permute as well.
6547 2018-05-11 Martin Liska <mliska@suse.cz>
6550 * doc/extend.texi: Document LLVM style format for no_sanitize
6553 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6555 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
6556 mode_supports_vsx_dform_quad to mode_supports_dq_form.
6557 (mode_supports_vsx_dform_quad): Likewise.
6558 (mode_supports_vmx_dform): Move these functions to be next to the
6559 other mode_supports functions.
6560 (mode_supports_dq_form): Likewise.
6561 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
6562 mode_supports_dq_form.
6563 (reg_offset_addressing_ok_p): Likewise.
6564 (offsettable_ok_by_alignment): Likewise.
6565 (rs6000_legitimate_offset_address_p): Likewise.
6566 (legitimate_lo_sum_address_p): Likewise.
6567 (rs6000_legitimize_address): Likewise.
6568 (rs6000_legitimize_reload_address): Likewise.
6569 (rs6000_secondary_reload_inner): Likewise.
6570 (rs6000_preferred_reload_class): Likewise.
6571 (rs6000_output_move_128bit): Likewise.
6573 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6575 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
6576 Generate SImode target register for null target.
6577 <case IX86_BUILTIN_XGETBV>: Ditto.
6578 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
6579 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
6581 2018-05-10 Carl Love <cel@us.ibm.com>
6583 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
6584 dcbtt and dcbtstt if operands[2] is 0.
6586 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
6589 * config/i386/sse.md (usadv64qi): New expander.
6591 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
6593 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
6594 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
6595 -maltivec=be support.
6596 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
6597 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
6598 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
6599 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
6600 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
6601 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
6602 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
6603 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
6604 altivec_vsumsws): Adjust.
6605 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
6606 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
6607 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
6609 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
6610 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
6611 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
6612 (altivec_lve<VI_char>x): Delete expand.
6613 (*altivec_lve<VI_char>x_internal): Rename to...
6614 (altivec_lve<VI_char>x): ... this.
6615 (altivec_lvxl_<mode>): Delete expand.
6616 (*altivec_lvxl_<mode>_internal): Rename to ...
6617 (altivec_lvxl_<mode>): ... this.
6618 (altivec_stvxl_<mode>): Delete expand.
6619 (*altivec_stvxl_<mode>_internal): Rename to ...
6620 (altivec_stvxl_<mode>): ... this.
6621 (altivec_stve<VI_char>x): Delete expand.
6622 (*altivec_stve<VI_char>x_internal): Rename to ...
6623 (altivec_stve<VI_char>x): ... this.
6624 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
6625 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
6626 reduc_plus_scal_<mode>): Adjust.
6627 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
6629 (rs6000_cpu_cpp_builtins): Adjust.
6630 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
6631 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
6632 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
6633 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
6634 -maltivec=be support.
6635 (rs6000_split_vec_extract_var): Adjust.
6636 (rs6000_split_v4si_init): Adjust.
6637 (swap_selector_for_mode): Delete.
6638 (altivec_expand_lvx_be, altivec_expand_stvx_be,
6639 altivec_expand_stvex_be): Delete.
6640 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
6641 -maltivec=be support.
6642 (rs6000_gimple_fold_builtin): Ditto.
6643 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
6645 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
6646 (TARGET_DIRECT_MOVE_64BIT): Adjust.
6647 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
6648 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
6649 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
6650 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
6651 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
6652 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
6653 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
6654 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
6655 anonymous split): Adjust.
6656 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
6657 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
6659 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
6661 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
6662 when --with-gxx-include-dir is also specified.
6663 * configure: Regenerate.
6665 2018-05-09 Jim Wilson <jimw@sifive.com>
6668 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
6669 * config/riscv/t-withmultilib: New.
6670 * config/riscv/withmultilib.h: New.
6671 * doc/install.texi: Document RISC-V --with-multilib-list support.
6673 2018-05-09 Richard Biener <rguenther@suse.de>
6675 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
6677 (vect_bb_vectorization_profitable_p): Adjust. Compute
6678 actual scalar cost using the cost vector and the add_stmt_cost
6681 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6683 PR rtl-optimization/85645
6684 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
6685 in the REG_CFA_REGISTER note for LR, don't leave it empty.
6687 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6689 PR rtl-optimization/85645
6690 * shrink-wrap.c (spread_components): Return a boolean saying if
6691 anything was changed.
6692 (try_shrink_wrapping_separate): Iterate spread_components until
6693 nothing changes anymore.
6695 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6697 PR rtl-optimization/85645
6698 * regrename.c (build_def_use): Also kill the chains that include the
6699 destination of a REG_CFA_REGISTER note.
6701 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
6703 PR rtl-optimization/85645
6704 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
6705 insn that has a REG_CFA_REGISTER note.
6707 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
6709 * cfgexpand.c (expand_clobber): New function.
6710 (expand_gimple_stmt_1): Use it.
6711 * tree-vect-stmts.c (vect_clobber_variable): New function,
6713 (vectorizable_simd_clone_call): ...here.
6714 (vectorizable_store): Emit a clobber either side of an
6715 IFN_STORE_LANES sequence.
6716 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
6718 2018-05-09 Tom de Vries <tom@codesourcery.com>
6721 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
6722 (define_insn "trap_if_false"): Add exit after trap.
6724 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
6726 PR rtl-optimization/85638
6727 * bb-reorder.c: Include common/common-target.h.
6728 (create_forwarder_block): New function extracted from...
6729 (fix_up_crossing_landing_pad): ...here. Rename into...
6730 (dw2_fix_up_crossing_landing_pad): ...this.
6731 (sjlj_fix_up_crossing_landing_pad): New function.
6732 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
6733 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
6734 from both partitions and exit the loop after one iteration.
6736 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6739 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6741 (Basic PowerPC Built-in Functions): The new name of the
6742 subsection previously known as "PowerPC Built-in Functions".
6743 (Basic PowerPC Built-in Functions Available on all Configurations):
6745 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6747 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6749 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6751 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6754 2018-05-08 Jim Wilson <jimw@sifive.com>
6756 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
6757 (LD_EMUL_SUFFIX): New.
6758 (LINK_SPEC): Use it.
6760 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
6762 * doc/extend.texi (PowerPC Built-in Functions): Rename this
6764 (Basic PowerPC Built-in Functions): The new name of the
6765 subsection previously known as "PowerPC Built-in Functions".
6766 (Basic PowerPC Built-in Functions Available on all Configurations):
6768 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
6770 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
6772 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
6774 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
6777 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6780 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
6781 after cmpelim optimization.
6783 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
6785 * config.gcc: Support "goldmont".
6786 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
6787 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6789 * config/i386/i386.c (m_GOLDMONT): Define.
6790 (processor_target_table): Add "goldmont".
6791 (PTA_GOLDMONT): Define.
6792 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
6793 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
6794 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
6795 (fold_builtin_cpu): Add "goldmont".
6796 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
6797 (ix86_option_override_internal): Add "goldmont".
6798 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
6799 (processor_type): Add PROCESSOR_GOLDMONT.
6800 * config/i386/i386.md: Add CPU "glm".
6801 * config/i386/glm.md: New file.
6802 * config/i386/x86-tune.def: Add m_GOLDMONT.
6803 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
6805 2018-05-08 Jakub Jelinek <jakub@redhat.com>
6808 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
6810 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
6811 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
6812 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
6815 * config/i386/i386.c (ix86_fold_builtin): Handle
6816 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
6819 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
6820 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
6822 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
6825 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
6826 (check_arch): Likewise.
6827 (check_fpu): Return the result rather than printing it.
6828 (end arch): Fix operator precedence.
6829 (end cpu): Likewise.
6830 (END): Print the result from check_fpu.
6832 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6833 Alan Hayward <alan.hayward@arm.com>
6834 David Sherwood <david.sherwood@arm.com>
6836 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
6837 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
6838 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
6839 (*fcmuo<mode>_and): New patterns.
6841 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
6843 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
6844 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
6845 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
6846 (cmp_op, sve_imm_con): New code attributes.
6847 (SVE_COND_INT_CMP, imm_con): Delete.
6848 (cmp_op): Remove above unspecs from int attribute.
6849 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
6851 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
6852 comparison-specific unspecs.
6853 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
6854 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
6855 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
6856 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
6857 (*vec_fcm<cmp_op><mode>): Rename to...
6858 (*fcm<cmp_op><mode>): ...this and adjust likewise.
6859 (*vec_fcmuo<mode>): Rename to...
6860 (*fcmuo<mode>): ...this and adjust likewise.
6861 (*pred_fcm<cmp_op><mode>): New pattern.
6862 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
6863 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
6865 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
6867 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
6868 (aarch64_emit_sve_predicated_cond): New function.
6869 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
6870 (aarch64_emit_unspec_cond_or): Replace with...
6871 (aarch64_emit_sve_or_conds): ...this new function. Use
6872 aarch64_emit_sve_ptrue_op for the individual comparisons and
6873 aarch64_emit_binop to OR them together.
6874 (aarch64_emit_inverted_unspec_cond): Replace with...
6875 (aarch64_emit_sve_inverted_cond): ...this new function. Use
6876 aarch64_emit_sve_ptrue_op for the comparison and
6877 aarch64_emit_unop to invert the result.
6878 (aarch64_expand_sve_vec_cmp_float): Update after the above
6879 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
6881 2018-05-07 Nathan Sidwell <nathan@acm.org>
6883 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
6884 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
6885 (Backwards Compatibility): Likewise.
6887 2018-05-07 Luis Machado <luis.machado@linaro.org>
6891 2018-05-07 Luis Machado <luis.machado@linaro.org>
6893 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6894 <prefetch_dynamic_strides>: New const bool field.
6895 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6896 prefetch_dynamic_strides.
6897 (exynosm1_prefetch_tune): Likewise.
6898 (thunderxt88_prefetch_tune): Likewise.
6899 (thunderx_prefetch_tune): Likewise.
6900 (thunderx2t99_prefetch_tune): Likewise.
6901 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6903 (aarch64_override_options_internal): Update to set
6904 PARAM_PREFETCH_DYNAMIC_STRIDES.
6905 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6906 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6907 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6908 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6909 prefetch-dynamic-strides setting.
6911 2018-05-07 Luis Machado <luis.machado@linaro.org>
6913 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6914 <minimum_stride>: New const int field.
6915 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6916 minimum_stride field.
6917 (exynosm1_prefetch_tune): Likewise.
6918 (thunderxt88_prefetch_tune): Likewise.
6919 (thunderx_prefetch_tune): Likewise.
6920 (thunderx2t99_prefetch_tune): Likewise.
6921 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6922 (aarch64_override_options_internal): Update to set
6923 PARAM_PREFETCH_MINIMUM_STRIDE.
6924 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6925 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6926 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6927 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6928 stride is constant and is below the minimum stride threshold.
6930 2018-05-07 Luis Machado <luis.machado@linaro.org>
6932 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
6935 2018-05-07 Luis Machado <luis.machado@linaro.org>
6937 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6938 <prefetch_dynamic_strides>: New const bool field.
6939 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6940 prefetch_dynamic_strides.
6941 (exynosm1_prefetch_tune): Likewise.
6942 (thunderxt88_prefetch_tune): Likewise.
6943 (thunderx_prefetch_tune): Likewise.
6944 (thunderx2t99_prefetch_tune): Likewise.
6945 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
6947 (aarch64_override_options_internal): Update to set
6948 PARAM_PREFETCH_DYNAMIC_STRIDES.
6949 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6950 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6951 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6952 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6953 prefetch-dynamic-strides setting.
6955 2018-05-07 Luis Machado <luis.machado@linaro.org>
6957 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6958 <minimum_stride>: New const int field.
6959 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6960 minimum_stride field.
6961 (exynosm1_prefetch_tune): Likewise.
6962 (thunderxt88_prefetch_tune): Likewise.
6963 (thunderx_prefetch_tune): Likewise.
6964 (thunderx2t99_prefetch_tune): Likewise.
6965 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
6966 (aarch64_override_options_internal): Update to set
6967 PARAM_PREFETCH_MINIMUM_STRIDE.
6968 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6969 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6970 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6971 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6972 stride is constant and is below the minimum stride threshold.
6974 2018-05-06 Jakub Jelinek <jakub@redhat.com>
6977 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
6978 the type is addressable. Don't force op into register if it has
6981 2018-05-05 Roland McGrath <mcgrathr@google.com>
6984 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
6985 any section for which we don't know a specific type it should have,
6986 regardless of name. Previously this was done only for the exact
6987 names ".init_array", ".fini_array", and ".preinit_array".
6988 (default_elf_asm_named_section): Add comment about
6989 relationship with default_section_type_flags and SECTION_NOTYPE.
6990 (get_section): Don't consider it a type conflict if one side has
6991 SECTION_NOTYPE and the other doesn't, as long as neither has the
6992 SECTION_BSS et al used in the default_section_type_flags logic.
6994 2018-05-05 Tom de Vries <tom@codesourcery.com>
6997 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
6998 (workaround_barsyncs): New function.
6999 (nvptx_reorg): Use workaround_barsyncs.
7000 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7001 (define_expand "nvptx_membar_cta"): New define_expand.
7002 (define_insn "*nvptx_membar_cta"): New insn.
7004 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7006 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7007 To improve optimization opportunities.
7008 * builtin-types.def: The new needed builtin types for the above.
7010 2018-05-04 Richard Biener <rguenther@suse.de>
7012 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7013 * gimple-ssa-store-merging.c
7014 (imm_store_chain_info::output_merged_store): Remove redundant create,
7015 release split_store vector contents on failure.
7016 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7017 scalar stmt vector on cache hit.
7019 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7021 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7023 * config.gcc (powerpc-xilinx-eabi*): Remove.
7024 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7026 (fusion_addis_mem_combo_load): Ditto.
7027 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7029 (rs6000_cpu_cpp_builtins): Ditto.
7030 * config/rs6000/rs6000-linux.c
7031 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7032 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7033 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7035 (rs6000_setup_reg_addr_masks): Ditto.
7036 (rs6000_init_hard_regno_mode_ok): Ditto.
7037 (rs6000_option_override_internal): Ditto.
7038 (legitimate_lo_sum_address_p): Ditto.
7039 (rs6000_legitimize_address): Ditto.
7040 (rs6000_legitimize_reload_address): Ditto.
7041 (rs6000_legitimate_address_p): Ditto.
7042 (abi_v4_pass_in_fpr): Ditto.
7043 (setup_incoming_varargs): Ditto.
7044 (rs6000_gimplify_va_arg): Ditto.
7045 (rs6000_split_multireg_move): Ditto.
7046 (rs6000_savres_strategy): Ditto.
7047 (rs6000_emit_prologue_components): Ditto.
7048 (rs6000_emit_epilogue_components): Ditto.
7049 (rs6000_emit_prologue): Ditto.
7050 (rs6000_emit_epilogue): Ditto.
7051 (rs6000_elf_file_end): Ditto.
7052 (rs6000_function_value): Ditto.
7053 (rs6000_libcall_value): Ditto.
7054 * config/rs6000/rs6000.h: Ditto.
7055 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7056 (TARGET_MINMAX): ... this. New.
7057 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7058 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7059 (*movsi_internal1_single): Delete.
7060 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7061 mfpu=, mxilinx-fpu): Delete.
7062 * config/rs6000/singlefp.h: Delete.
7063 * config/rs6000/sysv4.h: Remove Xilinx FP support.
7064 * config/rs6000/t-rs6000: Ditto.
7065 * config/rs6000/t-xilinx: Delete.
7066 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
7067 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
7068 (VStype_simple): Delete.
7069 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
7070 * config/rs6000/xfpu.h: Delete.
7071 * config/rs6000/xfpu.md: Delete.
7072 * config/rs6000/xilinx.h: Delete.
7073 * config/rs6000/xilinx.opt: Delete.
7074 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
7075 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
7077 2018-05-04 Tom de Vries <tom@codesourcery.com>
7080 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
7083 2018-05-04 Richard Biener <rguenther@suse.de>
7086 * tree-complex.c (update_complex_assignment): We are always in SSA form.
7087 (expand_complex_div_wide): Likewise.
7088 (expand_complex_operations_1): Likewise.
7089 (expand_complex_libcall): Preserve EH info of the original stmt.
7090 (tree_lower_complex): Handle removed blocks.
7091 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
7092 on complex multiplication and division libcall builtins.
7094 2018-05-04 Richard Biener <rguenther@suse.de>
7097 * fold-const.c (negate_expr_p): Restrict negation of operand
7098 zero of a division to when we know that can happen without
7100 (fold_negate_expr_1): Likewise.
7102 2018-05-04 Jakub Jelinek <jakub@redhat.com>
7105 * real.h (real_nextafter): Declare.
7106 * real.c (real_nextafter): New function.
7107 * fold-const-call.c (fold_const_nextafter): New function.
7108 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
7109 CASE_CFN_NEXTTOWARD.
7110 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
7111 even when arg1_mode is different from arg0_mode.
7113 2018-05-03 Nathan Sidwell <nathan@acm.org>
7115 * doc/extend.texi (Deprecated Features): Remove
7117 (Backwards Compatibility): Likewise.
7118 * doc/invoke.texi (C++ Language Options): Likewise.
7119 (C++ Dialect Options): Likewise.
7121 2018-05-03 Jakub Jelinek <jakub@redhat.com>
7124 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
7125 _mm512_mask_mullox_epi64): New intrinsics.
7127 2018-05-03 Tom de Vries <tom@codesourcery.com>
7130 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7131 dump files): Add offload-tree.
7133 2018-05-03 Richard Biener <rguenther@suse.de>
7135 PR tree-optimization/85615
7136 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
7137 to loops not nested in BBs loop father to avoid creating multi-entry
7140 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7142 PR tree-optimization/70291
7143 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
7144 arguments. Change return type to tree. Emit libcall as a new
7145 statement rather than replacing existing one when inplace_p is true.
7146 (expand_complex_multiplication_components): New function.
7147 (expand_complex_multiplication): Expand floating-point complex
7148 multiplication using the above.
7149 (expand_complex_division): Rename inner_type parameter to type.
7150 Update expand_complex_libcall call-site.
7151 (expand_complex_operations_1): Update expand_complex_multiplication
7152 and expand_complex_division call-sites.
7154 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7157 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7158 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7159 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
7160 the highest significant bit of the shift count mask is clear. In
7161 check whether and[sq]i3 is needed verify that all significant bits
7162 of the shift count other than the highest are set.
7164 2018-05-02 Tom de Vries <tom@codesourcery.com>
7167 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
7168 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
7169 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
7170 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
7171 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
7172 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
7173 __builtin_goacc_parlevel_size.
7175 2018-05-02 Richard Biener <rguenther@suse.de>
7177 PR tree-optimization/85597
7178 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
7179 do not use split vect_get_vec_defs call but call vect_get_slp_defs
7182 2018-05-02 Tom de Vries <tom@codesourcery.com>
7185 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7186 dump files): Add ltrans-tree.
7188 2018-05-02 Tom de Vries <tom@codesourcery.com>
7191 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
7192 dump files): Add wpa-ipa.
7194 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
7196 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
7197 powerpc*-*-linux*paired* target.
7198 * config/rs6000/750cl.h: Delete.
7199 * config/rs6000/paired.h: Delete.
7200 * config/rs6000/paired.md: Delete.
7201 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
7203 * config/rs6000/rs6000-builtin.def: Remove paired float support.
7204 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
7205 comment. Remove paired float support.
7206 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
7207 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7209 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
7210 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
7212 * config/rs6000/rs6000.c: Remove paired float support.
7213 (paired_expand_vector_init, paired_expand_vector_move,
7214 paired_emit_vector_compare, paired_emit_vector_cond_expr,
7215 (paired_expand_lv_builtin, paired_expand_stv_builtin,
7216 paired_expand_builtin, paired_expand_predicate_builtin,
7217 paired_init_builtins): Delete.
7218 * config/rs6000/rs6000.h: Remove paired float support.
7219 * config/rs6000/rs6000.md: Remove paired float support.
7220 (move_from_CR_ov_bit): Delete.
7221 * config/rs6000/rs6000.opt (mpaired): Delete.
7222 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
7223 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
7225 2018-05-02 Richard Biener <rguenther@suse.de>
7228 * gimplify.c (gimplify_save_expr): When in SSA form allow
7229 SAVE_EXPRs to compute to SSA vars.
7231 2018-05-02 Jakub Jelinek <jakub@redhat.com>
7234 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
7235 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
7236 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
7237 clobber operands[2], instead use a new pseudo. Formatting fixes.
7239 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
7241 PR tree-optimization/85586
7242 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
7243 exit early for statements in the same group if the accesses are
7246 2018-05-02 Tom de Vries <tom@codesourcery.com>
7249 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
7252 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
7254 PR tree-optimization/85143
7255 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
7257 2018-05-01 Tom de Vries <tom@codesourcery.com>
7260 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
7261 not found" error message.
7263 2018-05-01 Tom de Vries <tom@codesourcery.com>
7266 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
7267 * vec.c (test_ordered_remove_if): New function.
7268 (vec_c_tests): Call test_ordered_remove_if.
7269 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
7270 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
7271 * tree-vect-patterns.c (vect_pattern_recog_1): Use
7272 VEC_ORDERED_REMOVE_IF.
7274 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7276 PR tree-optimization/82665
7277 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
7278 pointer subtraction where arguments come from a memchr call.
7280 2018-05-01 Jakub Jelinek <jakub@redhat.com>
7282 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
7283 --push-state --as-needed and --pop-state instead of --as-needed and
7284 --no-as-needed if ld supports it.
7285 * configure: Regenerated.
7288 * doc/install.texi2html: Replace _002d with - and _002a with * in
7289 generated html files using sed.
7291 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7294 * gcc-rich-location.c (blank_line_before_p): New function.
7295 (use_new_line): New function.
7296 (gcc_rich_location::add_fixit_insert_formatted): New function.
7297 * gcc-rich-location.h
7298 (gcc_rich_location::add_fixit_insert_formatted): New function.
7300 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7302 * selftest.c (assert_streq): Rename "expected" and "actual" to
7303 "val1" and "val2". Extend NULL-handling to cover both inputs
7304 symmetrically, while still requiring both to be non-NULL for a pass.
7305 * selftest.h (assert_streq): Rename "expected" and "actual" to
7307 (ASSERT_EQ): Likewise.
7308 (ASSERT_EQ_AT): Likewise.
7309 (ASSERT_KNOWN_EQ): Likewise.
7310 (ASSERT_KNOWN_EQ_AT): Likewise.
7311 (ASSERT_NE): Likewise.
7312 (ASSERT_MAYBE_NE): Likewise.
7313 (ASSERT_MAYBE_NE_AT): Likewise.
7314 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
7315 the assertion to pass.
7316 (ASSERT_STREQ_AT): Likewise.
7318 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7320 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
7321 interaction with -pie.
7323 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7325 * selftest.h: Fix alphabetization of per-source-file selftest
7328 2018-04-30 Jason Merrill <jason@redhat.com>
7330 PR c++/61982 - dead stores to destroyed objects.
7331 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7334 2018-04-30 Jason Merrill <jason@redhat.com>
7336 * tree.c (build_clobber): New.
7337 * tree.h: Declare it.
7338 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7340 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7342 * diagnostic-show-locus.c (layout::layout): Update for
7343 location_get_source_line returning a char_span.
7344 (struct char_span): Move to input.h.
7345 (struct correction): Update for fields in char_span becoming
7347 (struct source_line): Update for location_get_source_line
7348 returning a char_span.
7349 (layout::print_line): Likewise.
7350 * edit-context.c (edited_file::print_content): Likewise.
7351 (edited_file::print_diff_hunk): Likewise.
7352 (edited_file::print_run_of_changed_lines): Likewise.
7353 (edited_file::get_num_lines): Likewise.
7354 (edited_line::edited_line): Likewise.
7355 * final.c (asm_show_source): Likewise.
7356 * input.c (location_get_source_line): Convert return type
7357 from const char * to char_span, losing the final "line_len"
7359 (dump_location_info): Update for the above.
7360 (get_substring_ranges_for_loc): Likewise. Use a char_span
7361 when handling the literal within the line.
7362 (test_reading_source_line): Update for location_get_source_line
7363 returning a char_span.
7364 * input.h (class char_span): Move here from
7365 diagnostic-show-locus.c, converting from a struct to a class.
7366 Make data members private.
7367 (char_span::operator bool): New.
7368 (char_span::length): New.
7369 (char_span::get_buffer): New.
7370 (char_span::operator[]): New.
7371 (char_span::subspan): Make const.
7372 (char_span::xstrdup): New.
7373 (location_get_source_line): Convert return type from const char *
7374 to char_span, losing the final "line_size" param.
7376 2018-04-30 Jan Hubicka <jh@suse.cz>
7378 * lto-wrapper.c (ltrans_priorities): New static var.
7379 (cmp_priority): New.
7380 (run_gcc): Read priorities and if doing parallel build order
7381 the Makefile by them.
7383 2018-04-30 David Malcolm <dmalcolm@redhat.com>
7385 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
7387 2018-04-30 Richard Biener <rguenther@suse.de>
7389 * tree-cfg.c (verify_address): Remove base argument, add
7390 flag whether to check TREE_ADDRESSABLE and do that.
7391 (verify_expr): Remove.
7392 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7393 (verify_gimple_assign_single): Likewise.
7394 (verify_gimple_switch): Likewise.
7395 (verify_expr_location_1): Dereference tp once. Add (disabled)
7396 piece from verify_expr.
7397 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
7399 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7401 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
7403 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7405 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
7406 (small_data_pattern): Likewise.
7407 (arc_rewrite_small_data): Likewise.
7408 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
7409 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
7410 (get_symbol_alignment): New function.
7411 (legitimate_small_data_address_p): Likewise.
7412 (legitimate_scaled_address): Update, call
7413 legitimate_small_data_address_p.
7414 (output_sdata): New static variable.
7415 (arc_print_operand): Update how we handle small data operands.
7416 (arc_print_operand_address): Likewise.
7417 (arc_legitimate_address_p): Update, use
7418 legitimate_small_data_address_p.
7419 (arc_rewrite_small_data_p): Remove.
7420 (arc_rewrite_small_data_1): Likewise.
7421 (arc_rewrite_small_data): Likewise.
7422 (small_data_pattern): Likewise.
7423 (compact_sda_memory_operand): Update to use
7424 legitimate_small_data_address_p and get_symbol_alignment.
7425 (prepare_move_operands): Don't rewite sdata pattern.
7426 (prepare_extend_operands): Remove.
7427 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
7429 (zero_extendqisi2): Likewise.
7430 (zero_extendhisi2): Likewise.
7431 (extendqihi2): Likewise.
7432 (extendqisi2): Likewise.
7433 (extendhisi2): Likewise.
7437 * config/arc/constraints.md (Usd): Change it to memory constraint.
7439 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
7441 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
7442 as source of std instructions.
7443 * config/arc/arc.md (movsi_insn): Update pattern predicate to
7444 allow 6-bit constants as source for store instructions.
7445 (movdi_insn): Update instruction pattern to allow 6-bit constants
7446 as source for store instructions.
7448 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
7450 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
7452 2018-04-30 Nathan Sidwell <nathan@acm.org>
7453 Sandra Loosemore <sandra@codesourcery.com>
7455 * dumpfile.c (dump_open): Allow '-' for stdout.
7456 * doc/invoke.texi (Developer Options): Document dump filename
7457 determination early. Document stdin/stdout selection.
7459 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
7461 Microblaze Target: PIC data text relative
7463 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
7464 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
7466 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
7467 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
7468 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
7469 New addressing mode for data-text relative position indepenedent code.
7470 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
7471 'ADDRESS_SYMBOLIC_TXT_REL'.
7472 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
7473 (microblaze_legitimate_pic_operand): Exclude function calls from
7474 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
7475 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
7477 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7478 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
7479 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
7480 for 'address + offset'.
7481 (microblaze_expand_prologue): Add new function prologue call for
7483 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
7484 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
7485 table in case of TARGET_PIC_DATA_TEXT_REL.
7486 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
7487 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
7488 Add new macros 'UNSPEC_TEXT',
7489 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
7490 + exclude function calls from 'UNSPEC_PLT' in case of data text
7492 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7493 new target hook for generating address diff vector tables in case of
7495 * doc/tm.texi : Regenerate.
7496 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
7497 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
7498 of addr diff vector generation.
7499 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
7500 target hook definition.
7501 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
7502 Add default function for generate_pic_addr_diff_vec -> flag_pic.
7503 * doc/invoke.texi (Add new pic option): Add new microblaze pic
7504 option for data text relative.
7506 2018-04-30 Richard Biener <rguenther@suse.de>
7508 * tree-chrec.h (evolution_function_is_constant_p): Remove
7510 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
7512 2018-04-30 Richard Biener <rguenther@suse.de>
7515 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
7517 2018-04-30 Richard Biener <rguenther@suse.de>
7519 PR tree-optimization/28364
7520 PR tree-optimization/85275
7521 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
7522 copying first exit test.
7524 2018-04-28 Mark Wielaard <mark@klomp.org>
7526 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
7528 (dwarf_AT): Handle DW_AT_addr_base.
7529 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
7531 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
7534 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
7535 (*ashl<dwi>3_doubleword_mask_1): Ditto.
7536 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
7537 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
7539 2018-04-28 Richard Biener <rguenther@suse.de>
7541 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
7542 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
7543 to reflect use. Only add interesting stmts.
7545 2018-04-27 Martin Jambor <mjambor@suse.cz>
7548 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
7549 the jump function allows for passing through aggregate values.
7551 2018-04-27 David Malcolm <dmalcolm@redhat.com>
7553 * input.h (in_system_header_at): Convert from macro to inline
7555 (from_macro_expansion_at): Likewise.
7556 (from_macro_definition_at): Likewise.
7558 2018-04-27 Jeff Law <law@redhat.com>
7560 * config.gcc: Mark tile* targets as deprecated/obsolete.
7562 2018-04-27 Richard Biener <rguenther@suse.de>
7564 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
7567 2018-04-27 Richard Biener <rguenther@suse.de>
7569 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
7571 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
7573 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
7574 with Yd constraint. Set "preferred_for_speed" attribute from
7575 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
7577 (*movdi_internal): Ditto.
7578 (movti_interunit splitters): Remove
7579 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
7580 (movdi_interunit splitters): Ditto.
7581 * config/i386/constraints.md (Ye): Remove.
7582 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
7584 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7587 * config/aarch64/constraints.md (Usg): Limit to 31.
7590 2018-04-27 Jakub Jelinek <jakub@redhat.com>
7592 PR tree-optimization/85529
7593 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
7594 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
7595 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
7596 zero extension or masking of the MSB bit.
7597 (optimize_range_tests): Add FIRST_BB argument, pass it through
7598 to optimize_range_tests_var_bound.
7599 (maybe_optimize_range_tests, reassociate_bb): Adjust
7600 optimize_range_tests callers.
7602 2018-04-26 Richard Biener <rguenther@suse.de>
7603 Jakub Jelinek <jakub@redhat.com>
7605 * cgraph.h (symbol_table): Just declare debug method here.
7606 * symtab.c (symbol_table::debug): Define.
7608 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
7610 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
7612 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
7614 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
7615 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
7616 (*movdi_internal): Substitute Yi and Yj constraint with x
7617 and Ym and Yn constraint with y constraint. Update "isa"
7618 attribute and set "preferred_for_speed" attribute from
7619 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
7620 (*movsi_internal): Ditto.
7621 (*movdf_internal): Ditto.
7622 (*movsf_internal): Ditto.
7623 (*zero_extendsidi2): Ditto.
7624 * config/i386/sse.md (vec_set<mode>_0): Ditto.
7625 (sse2_loadld): Ditto.
7626 (*vec_extract<ssevecmodelower>_0): Ditto.
7627 (*vec_extractv4si_0_zext_sse4): Ditto.
7628 (vec_concatv2di): Ditto.
7629 (*vec_dup<mode>): Ditto.
7630 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
7631 * config/i386/constraints.md (Yi): Remove.
7636 2018-04-26 Nathan Sidwell <nathan@acm.org>
7638 * dumpfile.c (dump_open): New.
7639 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
7640 (dump_finish): Detect stdio/stderr by value not name.
7642 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
7644 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
7646 2018-04-26 Tom de Vries <tom@codesourcery.com>
7649 * config/nvptx/nvptx.c (verify_neutering_jumps)
7650 (verify_neutering_labels): New function
7651 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
7653 2018-04-26 Tom de Vries <tom@codesourcery.com>
7656 * config/nvptx/nvptx.c (needs_neutering_p): New function.
7657 (nvptx_single): Use needs_neutering_p to skip over insns that do not
7660 2018-04-26 Richard Biener <rguenther@suse.de>
7661 Tom de Vries <tom@codesourcery.com>
7664 * lto-streamer-out.c (output_function): Fixup loops if required to match
7665 discovery done in the reader.
7667 2018-04-26 Richard Biener <rguenther@suse.de>
7669 PR tree-optimization/85116
7670 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
7671 have a loop exit from the single latch predecessor. Remove
7672 case of header with just condition.
7673 (ch_base::copy_headers): Exclude infinite loops from any
7675 (pass_ch::execute): Record exits.
7677 2018-04-26 Richard Biener <rguenther@suse.de>
7679 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
7680 prologue cost vector and pass it to vect_get_load_cost.
7681 (vect_get_peeling_costs_all_drs): Likewise.
7682 (vect_peeling_hash_get_lowest_cost): Likewise.
7683 (vect_enhance_data_refs_alignment): Likewise.
7685 2018-04-26 Richard Biener <rguenther@suse.de>
7688 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
7689 checking of integer<->pointer conversions.
7690 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
7691 sign-/zero-extending pointer types.
7692 (expand_omp_for_static_chunk): Likewise.
7694 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
7695 Jean Lee <xiaoyur347@gmail.com>
7697 * config/mips/mips.c (mips_asan_shadow_offset): New function.
7698 (TARGET_ASAN_SHADOW_OFFSET): Define.
7699 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
7700 true for -fsanitize=address.
7702 2018-04-25 Mark Wielaard <mark@klomp.org>
7704 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
7707 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7709 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
7710 than "alu", remove explicit "memory" and "imm_disp" attributes.
7711 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
7714 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
7715 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
7716 gen_lowpart_no_emit.
7718 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
7721 * config/i386/i386.c (ix86_expand_builtin): Change memory
7722 operand to XI, extend p0 to Pmode.
7723 * config/i386/i386.md: Change unspec volatile and operand
7724 1 mode to XI, change operand 0 mode to P.
7726 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7728 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
7729 GET_MODE_MASK before any checking.
7730 (nds32_can_use_bset_p): Likewise.
7731 (nds32_can_use_btgl_p): Likewise.
7733 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7735 * config/nds32/nds32-doubleword.md: New define_split pattern for
7736 illegal register number.
7738 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7740 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
7742 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
7744 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
7746 2018-04-25 Richard Biener <rguenther@suse.de>
7748 * lto-streamer.h (LTO_major_version): Bump to 8.
7750 2018-04-25 Jakub Jelinek <jakub@redhat.com>
7752 * BASE-VER: Set to 9.0.0.
7754 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
7756 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
7757 in __abskf2 and __powikf2.
7759 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7762 * config/aarch64/constraints.md (Usg, Usj): New constraints.
7763 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
7764 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
7765 Use the above on operand 2. Reindent.
7766 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
7768 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7771 * common/config/i386/i386-common.c (ix86_handle_option): Don't
7773 * config/i386/i386.opt (mcet): Removed.
7774 * doc/install.texi: Remove -mcet documentation.
7775 * doc/invoke.texi: Likewise.
7777 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
7780 * doc/install.texi: Remove -mcet from bootstrap-cet.
7782 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7785 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
7786 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
7790 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
7791 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
7792 containing a CONST_VECTOR.
7794 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
7796 * doc/install.texi: Update newlib dependency for nvptx.
7798 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7801 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
7802 instead of INTVAL when shifting x left.
7804 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
7806 PR tree-optimization/85478
7807 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
7808 vect_grouped_store_supported for single element vectors.
7810 2018-04-24 Richard Biener <rguenther@suse.de>
7813 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
7814 load cost increase to the case of non-constant step.
7816 2018-04-24 Jakub Jelinek <jakub@redhat.com>
7819 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
7820 destination if any_malformed_asm.
7822 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
7825 * expr.c (store_field): In the bitfield case, if the value comes from
7826 a function call and is returned in registers by means of a PARALLEL,
7827 do not change the mode of the temporary unless BLKmode and VOIDmode.
7829 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
7831 PR rtl-optimization/85423
7832 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
7833 dependencies to debug insns when the previous insn is non-debug.
7835 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7837 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
7838 enums into a single definition.
7839 (fls): Fix predicates and printing.
7842 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
7844 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
7845 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
7846 and short u6 immediate.
7847 (check_if_valid_sleep_operand): Remove.
7848 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
7850 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7852 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
7853 flag_always_save_lp condition.
7854 * config/nds32/nds32.opt (malways-save-lp): New option.
7856 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7858 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
7859 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
7860 * config/nds32/nds32.h
7861 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
7862 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
7864 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7866 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
7867 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
7869 2018-04-22 Shiva Chen <shiva0217@gmail.com>
7870 Chung-Ju Wu <jasonwucj@gmail.com>
7872 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
7874 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
7875 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
7877 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7879 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
7881 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7883 * config/nds32/nds32-protos.h (nds32_data_alignment,
7884 nds32_local_alignment): Declare.
7885 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
7886 nds32_local_alignment): New functions.
7887 (TARGET_CONSTANT_ALIGNMENT): Define.
7888 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
7890 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7892 * config/nds32/nds32.c
7893 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
7894 (TARGET_MODES_TIEABLE_P): Likewise.
7896 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7898 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
7901 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
7902 Chung-Ju Wu <jasonwucj@gmail.com>
7904 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7905 for unaligned access.
7906 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7908 * config/nds32/nds32-intrinsic.md: Likewise.
7909 * config/nds32/nds32_intrinsic.h: Likewise.
7910 * config/nds32/nds32.h (nds32_builtins): Likewise.
7911 * config/nds32/nds32.opt (munaligned-access): New option.
7912 * config/nds32/nds32.c (nds32_asm_file_start): Display
7913 flag_unaligned_access status.
7915 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7917 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
7918 -mno-relax is present.
7919 * config/riscv/linux.h (LINK_SPEC): Ditto.
7921 2018-04-20 Martin Sebor <msebor@redhat.com>
7924 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
7926 (gimple_fold_builtin_stxcpy_chk): Same.
7927 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
7929 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
7932 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
7933 __powikf2 when long double is IEEE 128-bit.
7935 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
7937 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
7938 step to make sure stack always aligned.
7940 2018-04-20 Carl Love <cel@us.ibm.com>
7943 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
7944 size check for arg0.
7946 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
7947 Tom de Vries <tom@codesourcery.com>
7950 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
7951 Emit insns for calls too.
7952 (nvptx_find_par): Always look for worker-level predecessor insn.
7953 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
7955 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
7956 (nvptx_process_pars): Propagate frames for calls.
7958 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
7961 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
7963 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
7964 (ix86_handle_option): Don't handle OPT_mibt.
7965 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
7967 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
7969 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
7970 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
7971 (ix86_target_macros): Define __CET__ with flag_cf_protection
7972 for -fcf-protection.
7973 * config/i386/i386.c (isa2_opts): Remove -mibt.
7974 * config/i386/i386.h (TARGET_IBT): Removed.
7975 (TARGET_IBT_P): Likewise.
7976 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
7977 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
7978 * config/i386/i386.opt (mcet): Update help message.
7981 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
7982 -mcet as an alias for -mshstk.
7984 2018-04-20 Richard Biener <rguenther@suse.de>
7987 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
7988 complexity by forcing a single use of the multiply operand.
7990 2018-04-20 Martin Jambor <mjambor@suse.cz>
7993 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
7994 recursion dependency to only apply to non-clones.
7996 2018-04-20 Martin Jambor <mjambor@suse.cz>
7999 * ipa-cp.c (create_specialized_node): Check that clones of
8000 self-recursive edges exist during IPA-CP.
8002 2018-04-19 Toon Moene <toon@moene.org>
8004 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8007 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8009 PR tree-optimization/85467
8010 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8011 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8012 VECTOR_CST element to type.
8014 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8017 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8018 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8019 (builtin_longjmp): Likewise.
8020 (save_stack_nonlocal): New pattern.
8021 (restore_stack_nonlocal): Likewise.
8023 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8026 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8027 Replace ASM_OUTPUT_LABEL with fprintf.
8029 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8032 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8033 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8034 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8035 define __IBT__ and __SHSTK__ for -fcf-protection.
8036 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8038 (ix86_trampoline_init): Likewise.
8039 (x86_output_mi_thunk): Likewise.
8040 (ix86_notrack_prefixed_insn_p): Likewise.
8041 (ix86_option_override_internal): Don't disallow -fcf-protection.
8042 * config/i386/i386.md (rdssp<mode>): Also enable for
8044 (incssp<mode>): Likewise.
8045 (nop_endbr): Likewise.
8046 * config/i386/i386.opt (mcet): Change help message to built-in
8050 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8051 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8052 enable CET built-in functions.
8054 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8056 * common/config/i386/i386-common.c
8057 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8058 OPTION_MASK_ISA_MOVDIRI_UNSET,
8059 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8060 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8061 * config.gcc (movdirintrin.h): New header.
8062 * config/i386/cpuid.h (bit_MOVDIRI,
8063 bit_MOVDIR64B): New bits.
8064 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
8066 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
8067 (VOID, PVOID, PCVOID)): New function types.
8068 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
8069 __builtin_ia32_directstoreu_u64,
8070 __builtin_ia32_movdir64b): New builtins.
8071 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
8072 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
8074 (ix86_valid_target_attribute_inner_p): Ditto.
8075 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
8076 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
8077 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
8078 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
8079 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
8080 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
8081 (movdiri<mode>, movdir64b_<mode>): New.
8082 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
8083 * config/i386/immintrin.h: Include movdirintrin.h.
8084 * config/i386/movdirintrin.h: New file.
8085 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
8087 2018-04-19 Richard Biener <rguenther@suse.de>
8090 * cfg.c (clear_bb_flags): When loop state says we have
8091 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
8093 2018-04-19 Richard Biener <rguenther@suse.de>
8095 PR tree-optimization/84737
8096 * tree-vect-data-refs.c (vect_copy_ref_info): New function
8097 copying restrict info.
8098 (vect_setup_realignment): Use it.
8099 * tree-vectorizer.h (vect_copy_ref_info): Declare.
8100 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
8101 the first DR to all generated stores.
8102 (vectorizable_load): Likewise for loads.
8104 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8106 PR tree-optimization/85446
8107 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
8108 the integral and pointer types to have the same precision.
8110 * doc/install.texi: Document --disable-cet being the default and
8113 2018-04-18 Martin Liska <mliska@suse.cz>
8115 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
8118 2018-04-18 Martin Liska <mliska@suse.cz>
8121 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8124 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8125 arguments if they are comparable.
8127 2018-04-18 Martin Liska <mliska@suse.cz>
8130 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8133 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8136 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
8139 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
8140 ENDBR after calling __morestack.
8142 2018-04-18 David Malcolm <dmalcolm@redhat.com>
8145 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
8146 by using gcc_base_ver to generate a gcc_driver_version, and use
8147 it when generating GCC_DRIVER_NAME.
8148 * configure: Regenerate.
8150 2018-04-18 Jakub Jelinek <jakub@redhat.com>
8153 * config.gcc: Obsolete powerpc*-*-*spe*.
8155 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8158 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
8159 (stabstr_D): Change type of unum from unsigned int to
8160 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
8163 2018-04-17 Jim Wilson <jimw@sifive.com>
8166 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
8167 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
8168 Set arg_pointer_offset after using pretend_args_size.
8170 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8172 PR rtl-optimization/85431
8173 * dse.c (record_store): Ignore zero width stores.
8176 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
8177 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
8178 __builtin_stack_restore rather than after it.
8179 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
8180 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
8181 argument instead of virtual_dynamic_stack_rtx.
8183 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
8185 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
8187 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
8188 Add note to error message to explain internal mapping of overloaded
8189 built-in function name to non-overloaded built-in function name.
8190 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
8193 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8196 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
8197 where the inputs overlap with the output.
8199 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8201 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
8202 (=v, v) alternative and explicit "memory" attribute.
8203 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
8204 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8206 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
8207 "sselog1" type instead of "sselog".
8208 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
8209 "sselog". Remove explicit "memory" attribute.
8210 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
8211 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8213 (vec_extract_hi_v32hi): Merge all alternatives into one, use
8214 "sselog1" type instead of "sselog". Remove explicit "memory"
8216 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
8217 use "sselog1" type instead of "sselog". Remove explicit "memory"
8219 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
8220 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
8222 (vec_extract_hi_v64qi): Merge all alternatives into one, use
8223 "sselog1" type instead of "sselog". Remove explicit "memory"
8225 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
8226 use "sselog1" type instead of "sselog". Remove explicit "memory"
8230 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
8233 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
8236 2018-04-17 Martin Jambor <mjambor@suse.cz>
8239 * ipa-cp.c (create_specialized_node): Call
8240 expand_all_artificial_thunks if necessary.
8242 2018-04-17 Martin Liska <mliska@suse.cz>
8245 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
8246 in message, remote space in between '_G' and '('.
8248 2018-04-17 Jakub Jelinek <jakub@redhat.com>
8251 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
8252 avx512f_vmcmp<mode>3<round_saeonly_name>,
8253 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
8254 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
8255 avx512f_rndscale<mode><round_saeonly_name>,
8256 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
8257 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
8258 Use %<iptr>2 instead of %2 for -masm=intel.
8259 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
8260 avx512f_vcvttss2usi<round_saeonly_name>,
8261 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
8263 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
8264 avx512f_vcvttsd2usi<round_saeonly_name>,
8265 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
8266 Use %q1 instead of %1 for -masm=intel.
8267 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
8268 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
8269 of %3 for -masm=intel.
8270 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
8272 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
8274 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
8276 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
8277 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
8279 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
8281 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
8282 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
8283 %g1 and one with %0 and %1.
8284 (avx512er_vmrcp28<mode><round_saeonly_name>,
8285 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
8287 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
8288 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
8289 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
8290 of %0 and %{%4%} for -masm=intel.
8291 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
8292 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
8293 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
8294 order of %0 and %{%5%}%{z%} for -masm=intel.
8296 2018-04-17 Jan Hubicka <jh@suse.cz>
8299 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
8301 2018-04-17 Martin Liska <mliska@suse.cz>
8304 * multiple_target.c (create_dispatcher_calls): Set apostrophes
8305 for target_clone error message. Make default implementation
8306 clone to be a local declaration.
8307 (separate_attrs): Add new argument and check for an empty
8309 (expand_target_clones): Handle it.
8310 (ipa_target_clone): Make redirection just for target_clones
8313 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
8314 Tom de Vries <tom@codesourcery.com>
8317 * omp-expand.c (expand_oacc_for): Add dummy false branch for
8318 tiled basic blocks without omp continue statements.
8320 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
8323 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
8324 vec_extract expression as having side effects to make sure it gets
8327 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
8330 * config/i386/i386.c (get_builtin_code_for_version): Check
8333 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
8336 * gcc/config.gcc: Support "skylake".
8337 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
8339 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
8340 (processor_target_table): Add "skylake".
8341 (ix86_option_override_internal): Add "skylake".
8342 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
8343 PROCESSOR_CANNONLAKE.
8344 (get_builtin_code_for_version): Fix priority for
8345 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
8346 PROCESSOR_SKYLAKE-AVX512.
8347 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
8348 (processor_type): Add PROCESSOR_SKYLAKE.
8350 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
8351 Jason Merrill <jason@redhat.com>
8354 * convert.c (convert_to_integer_1): Use direct recursion for
8355 enumeral types and types with a precision less than the number
8356 of bits in their mode.
8358 2018-04-16 Julia Koval <julia.koval@intel.com>
8361 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
8362 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
8364 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
8367 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
8368 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8369 and -mno-direct-move.
8371 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
8374 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
8375 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
8376 Ensure negative shifts result in {0}.
8378 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
8380 PR rtl-optimization/79916
8381 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
8382 regs (if any) to define how to gnerate SD moves when LRA is in
8385 2018-04-13 Jakub Jelinek <jakub@redhat.com>
8387 PR rtl-optimization/85393
8388 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
8389 * except.c (expand_dw2_landing_pad_for_region): Make static.
8390 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
8391 a label and unconditional jump to old_bb, rather than
8392 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
8395 PR rtl-optimization/85376
8396 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
8397 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
8398 instead of a specific value.
8400 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8401 Bin Cheng <bin.cheng@arm.com>
8403 PR tree-optimization/82965
8404 PR tree-optimization/83991
8405 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
8406 by_profile_only parameter.
8407 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
8408 information if the loop was predicted to iterate too many times.
8409 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
8411 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
8414 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
8417 2018-04-13 Martin Liska <mliska@suse.cz>
8418 Jakub Jelinek <jakub@redhat.com>
8421 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
8422 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
8423 * builtins.c (expand_builtin_memory_copy_args): Use
8424 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
8425 handle dest_addr == pc_rtx.
8427 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
8430 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
8431 asked to not generate direct moves.
8432 (fix_trunc<mode>si2_stfiwx): Similar.
8433 (fix_trunc<mode>si2_internal): Similar.
8435 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8438 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
8439 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
8440 lookup if dest in some wider mode is known to be const0_rtx and
8441 if so, record permanent equivalence for it to be ZERO_EXTEND of
8442 the narrower mode destination.
8444 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8446 * lto-streamer-out.c (output_function): Revert 259346.
8447 * omp-expand.c (expand_oacc_for): Likewise.
8449 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
8451 PR rtl-optimization/85354
8452 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
8453 * sel-sched.c (sel_global_init): ... here.
8455 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
8458 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
8459 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8460 mode for PE-COFF targets.
8461 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
8462 (i386_pe_asm_lto_end): Likewise.
8463 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
8464 (TARGET_ASM_LTO_END): Likewise.
8465 * config/i386/winnt.c (saved_debug_info_level): New static variable.
8466 (i386_pe_asm_lto_start): New function.
8467 (i386_pe_asm_lto_end): Likewise.
8469 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
8470 Richard Biener <rguenther@suse.de>
8473 * lto-streamer-out.c (output_function): Fix CFG loop state before
8475 * omp-expand.c (expand_oacc_for): Handle calls to internal
8476 functions like regular functions.
8478 2018-04-12 Richard Biener <rguenther@suse.de>
8481 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
8482 for the early LTO debug to properly generate references to it
8483 during DIE emission. Do not re-use that for the skeleton for
8485 (dwarf2out_early_finish): Likewise.
8487 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8490 * config/i386/sse.md
8491 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
8492 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
8493 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8494 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
8495 and output is a reg, avoid creating invalid lowpart subreg, but
8496 instead split into a 512-bit move. Don't split if not AVX512VL,
8497 input is xmm16+ reg and output is a mem.
8498 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
8499 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
8500 xmm16+ reg and output is a mem.
8502 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8504 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
8505 also for flag_dwarf2_cfi_asm.
8507 2018-04-12 Jakub Jelinek <jakub@redhat.com>
8509 PR rtl-optimization/85342
8510 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
8511 a bool scalar var inside of the loop instead. Don't try to update
8512 recog_data.operand after failed apply_change_group.
8514 2018-04-12 Tom de Vries <tom@codesourcery.com>
8517 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
8518 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
8519 array with flexible array member as array without given dimension.
8520 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
8521 argument for undefined param to true.
8523 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
8526 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8527 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
8528 from PowerPC section.
8529 * config/rs6000/sysv4.opt (mcall-): Improve help text.
8530 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
8531 help text that is too long.
8532 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
8533 help text that is too long.
8534 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
8535 help text that is too long.
8537 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
8539 * config/alpha/alpha.md (stack_probe_internal): Rename
8540 from "probe_stack". Update all callers.
8542 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8544 PR rtl-optimization/84566
8545 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
8546 sched_macro_fuse_insns.
8548 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8551 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
8552 (compute_block_dependences): ... from here.
8554 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8556 PR tree-optimization/85331
8557 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
8558 from int to HOST_WIDE_INT.
8560 2018-04-11 Martin Jambor <mjambor@suse.cz>
8563 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
8564 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
8565 not the same as the source val.
8566 (cgraph_edge_brings_value_p): New parameter.
8567 (gather_edges_for_value): Pass destination value to
8568 cgraph_edge_brings_value_p.
8569 (perhaps_add_new_callers): Likewise.
8570 (get_info_about_necessary_edges): Likewise and exclude values brought
8571 only by self-recursive edges.
8572 (create_specialized_node): Redirect only clones of self-calling edges.
8573 (+self_recursive_pass_through_p): New function.
8574 (find_more_scalar_values_for_callers_subset): Use it.
8575 (find_aggregate_values_for_callers_subset): Likewise.
8576 (known_aggs_to_agg_replacement_list): Removed.
8577 (decide_whether_version_node): Re-calculate known constants for all
8578 remaining context clones.
8580 2018-04-11 Richard Biener <rguenther@suse.de>
8583 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
8584 from early DWARF output.
8585 (dwarf2out_early_finish): Output line info unconditionally into
8586 early DWARF and add reference to it.
8588 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8591 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
8592 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
8593 other than V2DFmode using iptr mode attribute.
8594 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
8596 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
8598 PR rtl-optimization/84659
8599 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
8601 2018-04-11 Jakub Jelinek <jakub@redhat.com>
8604 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
8606 (output_loc_list): Pass address of a dummy size variable even in the
8607 locview handling loop.
8608 (index_location_lists): Add comment on why skip_loc_list_entry can't
8611 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
8614 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
8617 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
8620 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
8621 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
8622 and -mstring-compare-inline-limit.
8624 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8627 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
8628 for stack clash protection in a register whenever we need it to be in
8631 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8633 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
8634 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
8636 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
8639 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
8641 (mlong-double-): Ditto.
8642 * config/rs6000/sysv4.opt (msdata=): Ditto.
8643 (mtls-size=): Ditto.
8645 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8647 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8648 erroneous entries for
8649 "vector int vec_ldl (int, long int *)", and
8650 "vector unsigned int vec_ldl (int, unsigned long int *)".
8651 Add comments and entries for
8652 "vector bool char vec_ldl (int, bool char *)",
8653 "vector bool short vec_ldl (int, bool short *)",
8654 "vector bool int vec_ldl (int, bool int *)",
8655 "vector bool long long vec_ldl (int, bool long long *)",
8656 "vector pixel vec_ldl (int, pixel *)",
8657 "vector long long vec_ldl (int, long long *)",
8658 "vector unsigned long long vec_ldl (int, unsigned long long *)".
8659 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
8660 type tree bool_long_long_type_node and correct definition of
8661 bool_V2DI_type_node to make reference to this new type tree.
8662 (rs6000_mangle_type): Replace erroneous reference to
8663 bool_long_type_node with bool_long_long_type_node.
8664 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
8665 comments to emphasize sign distinctions for char and int types and
8666 replace RS6000_BTI_bool_long constant with
8667 RS6000_BTI_bool_long_long constant. Also add comment to restrict
8668 use of RS6000_BTI_pixel.
8669 (bool_long_type_node): Remove this macro definition.
8670 (bool_long_long_type_node): New macro definition
8672 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8674 PR rtl-optimization/85300
8675 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
8676 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
8677 simplify_unary_operation fails.
8679 2018-04-10 Martin Liska <mliska@suse.cz>
8681 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
8682 cgraph_edge and ipa_ref.
8684 2018-04-10 Jakub Jelinek <jakub@redhat.com>
8688 * config/i386/sse.md
8689 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
8690 computation of the VEC_MERGE selector from mask.
8691 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
8692 Fix decoding of the VEC_MERGE selector into mask.
8694 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8696 PR tree-optimization/85286
8697 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8699 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
8701 * final.c (final_1): Set insn_last_address as well as
8702 insn_current_address.
8704 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8707 * explow.c (emit_stack_probe): Call validize_mem on memory location
8708 before passing it to gen_probe_stack. Create address operand and
8709 legitimize it for the probe_stack_address case.
8711 2018-04-09 Jan Hubicka <jh@suse.cz>
8714 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
8715 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
8716 * tree.c (free_lang_data_in_type): Fix handling of binfos;
8718 (free_lang_data): Rebuild type inheritance graph.
8720 2018-04-09 Martin Sebor <msebor@redhat.com>
8722 * invoke.texi (-finline-small-functions): Mention other optimization
8724 (-findirect-inlining, -fpartial-inlining): Same.
8725 (-finline-functions-called-once): Same.
8726 (-freorder-blocks-and-partition): Same.
8728 2018-04-09 Jan Hubicka <jh@suse.cz>
8731 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
8732 jumps; choose last target that matches the criteria (i.e.
8733 no partition changes for non-crossing jumps).
8734 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
8735 support for redirecting crossing jumps to non-crossing.
8737 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
8739 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
8740 also for naked functions.
8742 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
8744 * config/arc/arc.md (add_shift): New pattern.
8745 (add_shift2): Likewise.
8746 (sub_shift): Likewise.
8747 (sub_shift_cmp0_noout): Likewise.
8748 (compare_si_ashiftsi): Likewise.
8749 (xbfu_cmp0_noout): New combine pattern.
8750 (xbfu_cmp0"): Likewise.
8751 (movsi_set_cc_insn): Place the predicable variant first.
8752 (commutative_binary_cmp0_noout): Remove clobber.
8753 (commutative_binary_cmp0): New pattern.
8754 (noncommutative_binary_cmp0): Likewise.
8755 (noncommutative_binary_cmp0_noout): Likewise.
8756 (noncommutative_binary_comparison_result_used): Removed.
8757 (rsub_cmp0): New pattern.
8758 (rsub_cmp0_noout): Likewise.
8759 (extzvsi): Changed, keep only meaningful variants.
8760 (SQH, SEZ): New iterators.
8761 (SQH_postfix): New mode attribute.
8762 (SEZ_prefix): New code attribute.
8763 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
8764 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
8765 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
8767 (noncommutative_operator): Check the availability of barrel
8770 2018-04-09 Richard Biener <rguenther@suse.de>
8772 PR tree-optimization/85284
8773 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
8774 Only use the niter constraining form of simple_iv when the exit
8777 2018-04-09 Tom de Vries <tom@codesourcery.com>
8780 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
8781 (define_expand "*memory_barrier"): New define_expand.
8782 (define_insn "memory_barrier"): New insn.
8784 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8786 PR rtl-optimization/80463
8787 PR rtl-optimization/83972
8788 PR rtl-optimization/83480
8790 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
8791 correct producer for the insn.
8792 (tidy_control_flow): Fixup seqnos in case of debug insns.
8794 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8796 PR rtl-optimization/83913
8798 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
8799 different sched-times when merging exprs.
8801 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8803 PR rtl-optimization/83962
8805 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
8806 tidy_fallthru_edge and tidy_control_flow.
8808 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
8810 PR rtl-optimization/83530
8812 * sel-sched.c (force_next_insn): New global variable.
8813 (remove_insn_for_debug): When force_next_insn is true, also leave only
8814 next insn in the ready list.
8815 (sel_sched_region): When the region wasn't scheduled, make another pass
8816 over it with force_next_insn set to 1.
8818 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8820 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
8822 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8823 for interrupt control.
8824 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8825 functions for interrupt control.
8826 * config/nds32/nds32-intrinsic.md: Likewise.
8827 * config/nds32/nds32_intrinsic.h: Likewise.
8828 * config/nds32/nds32.h (nds32_builtins): Likewise.
8830 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8832 * config/nds32/nds32.c (nds32_init_machine_status,
8833 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
8834 strict_aligned_p field.
8835 (nds32_expand_to_rtl_hook): New function.
8836 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8837 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
8839 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8840 Chung-Ju Wu <jasonwucj@gmail.com>
8842 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
8843 * config/nds32/nds32-n7.md: New file.
8844 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
8845 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
8847 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
8848 * config/nds32/nds32.md (pipeline_model): Add n7.
8849 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
8850 * config/nds32/pipelines.md: Include n7 settings.
8852 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8853 Chung-Ju Wu <jasonwucj@gmail.com>
8855 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
8856 * config/nds32/nds32-e8.md: New file.
8857 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
8858 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
8860 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
8861 * config/nds32/nds32.md (pipeline_model): Add e8.
8862 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
8863 * config/nds32/pipelines.md: Include e8 settings.
8865 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8866 Chung-Ju Wu <jasonwucj@gmail.com>
8868 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
8869 * config/nds32/nds32-n8.md: New file.
8870 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
8871 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
8873 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
8874 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
8875 * config/nds32/nds32.md (pipeline_model): Add n8.
8876 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
8877 * config/nds32/pipelines.md: Include n8 settings.
8879 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
8880 Chung-Ju Wu <jasonwucj@gmail.com>
8882 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
8883 * config/nds32/nds32-n9-2r1w.md: New file.
8884 * config/nds32/nds32-n9-3r2w.md: New file.
8885 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
8886 nds32_register_ports): New or modify for cpu n9.
8887 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
8889 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
8890 * config/nds32/nds32-utils.c: New file.
8891 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
8892 TARGET_MUL_SLOW): Define.
8893 * config/nds32/nds32.md (pipeline_model): New attribute.
8894 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
8895 New options that support cpu n9.
8896 * config/nds32/pipelines.md: Include n9 settings.
8897 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
8899 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
8901 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
8902 information if necessary.
8903 (output_cond_branch_compare_zero): Likewise.
8904 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
8905 (nds32_target_alignment): Refine for alignment.
8906 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
8907 (FUNCTION_BOUNDARY): Modify.
8908 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
8910 * config/nds32/nds32.opt (malways-align, malign-functions): New.
8912 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
8914 * config/nds32/constants.md (unspec_volatile_element): Add values for
8915 TLB operation and data prefetch.
8916 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
8917 functions for TLB operation and data prefetch.
8918 * config/nds32/nds32-intrinsic.md: Likewise.
8919 * config/nds32/nds32_intrinsic.h: Likewise.
8920 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
8921 (nds32_print_operand): Likewise.
8922 * config/nds32/nds32.h (nds32_builtins): Likewise.
8924 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
8925 Andrew Pinski <pinsika@gcc.gnu.org>
8928 * match.pd: Use constant_boolean_node of correct type instead of
8929 boolean_true_node or boolean_false_node for simplifying
8930 pointer comparisons to zero.
8932 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8934 PR tree-optimization/80021
8935 * tree.c (verify_type_variant): Make error call in verify_variant_match
8936 translatable and remove final full stop.
8938 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8940 * config/nds32/constants.md (unspec_volatile_element): Add
8941 UNSPEC_VOLATILE_EH_RETURN.
8942 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
8943 nds32_output_stack_pop): Support dwarf exception handling process.
8944 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
8945 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
8946 exception handling process.
8947 (nds32_compute_stack_frame): Likewise.
8948 (nds32_return_addr_rtx): Likewise.
8949 (nds32_initial_elimination_offset): Likewise.
8950 (nds32_expand_prologue): Likewise.
8951 (nds32_expand_epilogue): Likewise.
8952 (nds32_dynamic_chain_address): New function.
8953 * config/nds32/nds32.h (machine_function): Add fields for dwarf
8955 (DYNAMIC_CHAIN_ADDRESS): Define.
8956 (EH_RETURN_DATA_REGNO): Define.
8957 (EH_RETURN_STACKADJ_RTX): Define.
8958 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
8959 patterns for dwarf exception handling.
8961 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
8963 * config/nds32/nds32.h: Clean up obsolete macros.
8965 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8967 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8968 Add enum values for particular instructions.
8969 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8970 particular intrinsic functions.
8971 * config/nds32/nds32-intrinsic.md: Likewise.
8972 * config/nds32/nds32_intrinsic.h: Likewise.
8973 * config/nds32/nds32.h (nds32_builtins): Likewise.
8974 * config/nds32/nds32.md (type): Add pbsad and pbsada.
8975 (btst, ave): New patterns for particular instructions.
8977 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8979 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8980 Add enum values for atomic load/store and memory sync.
8981 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
8983 * config/nds32/nds32-intrinsic.md: Likewise.
8984 * config/nds32/nds32_intrinsic.h: Likewise.
8985 * config/nds32/nds32.h (nds32_builtins): Likewise.
8987 2018-04-07 Jakub Jelinek <jakub@redhat.com>
8989 PR tree-optimization/85257
8990 * fold-const.c (native_encode_vector): If not all elts could fit
8991 and off is -1, return 0 rather than offset.
8992 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
8993 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
8994 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
8995 adjust buffer in native_interpret_expr call.
8997 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
8999 * config/nds32/constants.md (unspec_volatile_element): Add cache
9000 control enum values.
9001 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9002 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9003 * config/nds32/nds32.c (nds32_cctl_names): New.
9004 (nds32_print_operand): Handle cache control register names.
9005 * config/nds32/nds32.h (nds32_builtins): New enum values.
9006 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9008 * config/nds32/nds32.md (type): Add mmu.
9009 * config/nds32/pipelines.md (simple_insn): Add mmu.
9011 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9013 * config/nds32/nds32.md (type): Remove call.
9014 * config/nds32/pipelines.md (simple_insn): Likewise.
9016 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9018 * config/nds32/constants.md (unspec_volatile_element): Add
9019 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9020 UNSPEC_VOLATILE_FMFCFG.
9021 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9022 description for fmfcfg and fmfcsr.
9023 (bdesc_1arg): Add fmtcsr.
9024 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9025 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9026 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9027 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9028 unspec_fmfcfg): New patterns.
9029 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9030 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9031 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9032 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9033 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9034 __nds32__fmfcfg): Define.
9036 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9038 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9039 intrinsic register names.
9040 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9041 intrinsic register enum values and macros.
9043 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9045 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9046 for load/store addressing form.
9047 (nds32_print_operand_address): Likewise.
9049 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9052 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9053 based on LABEL_REF. Remove useless assertion.
9054 (pic_address_needs_scratch): Fix formatting.
9055 (sparc_legitimize_pic_address): Minor tweaks.
9056 (sparc_delegitimize_address): Adjust assertion accordingly.
9057 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9058 into symbolic_operand.
9059 (movsi_high_pic_label_ref): Likewise.
9060 (movsi_lo_sum_pic_label_ref): Likewise.
9061 (movdi_pic_label_ref): Likewise.
9062 (movdi_high_pic_label_ref): Likewise.
9063 (movdi_lo_sum_pic_label_ref): Likewise.
9065 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
9067 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
9068 custom LIB_SPEC setup.
9070 2018-04-06 Ruslan Bukin <br@bsdpad.com>
9071 Kito Cheng <kito.cheng@gmail.com>
9073 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
9074 * config/riscv/freebsd.h: New.
9076 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9078 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
9079 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
9082 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9083 Kito Cheng <kito.cheng@gmail.com>
9085 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
9086 nds32_output_call, nds32_symbol_binds_local_p): New functions.
9087 * config/nds32/nds32-protos.h (nds32_output_call,
9088 nds32_output_return): Declare.
9089 * config/nds32/nds32.md: Refine all the call and return patterns.
9091 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9094 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
9095 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
9097 PR rtl-optimization/84872
9098 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
9099 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
9102 2018-04-06 Tamar Christina <tamar.christina@arm.com>
9104 * expr.c (copy_blkmode_to_reg): Revert 254862.
9105 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
9107 2018-04-06 Richard Biener <rguenther@suse.de>
9110 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
9111 after seeing a component reference with an adjacent field. Treat
9112 refs to arrays at struct end of external decls similar to
9113 refs to unconstrained commons.
9115 2018-04-06 Jakub Jelinek <jakub@redhat.com>
9118 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
9119 look through SAVE_EXPRs with non-side-effects argument. Adjust
9121 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
9124 2018-04-06 Richard Biener <rguenther@suse.de>
9127 * alias.c (find_base_term): New wrapper around find_base_term
9128 unwinding CSELIB_VAL_PTR changes.
9129 (find_base_term): Do not restore CSELIB_VAL_PTR during the
9132 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9134 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
9136 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
9137 constant definitions.
9138 ("nop"): lr 0,0 -> nopr r0
9139 ("nop_lr0", "nop_lr1"): New insn definitions.
9141 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
9143 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
9144 NDS32_V3PUSH_AVAILABLE_P macro.
9146 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
9147 Chung-Ju Wu <jasonwucj@gmail.com>
9149 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
9150 (nds32*-*-*): Add float and fpu_config into supported_defaults.
9151 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
9152 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
9153 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
9154 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
9155 * config/nds32/constraints.md: New constraints and checking for hard
9156 float configuration.
9157 * config/nds32/iterators.md: New mode iterator and attribute for hard
9158 float configuration.
9159 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
9161 * config/nds32/nds32-fpu.md: New file.
9162 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
9163 deal with hard float code generation.
9164 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
9166 (abi_type, float_reg_number): New enum type.
9167 * config/nds32/nds32-predicates.c: New predicates for hard float.
9168 * config/nds32/nds32-protos.h: Declare functions for hard float.
9169 * config/nds32/nds32.c: Implementation for hard float configuration.
9170 * config/nds32/nds32.h: Definitions for hard float configuration.
9171 * config/nds32/nds32.md: Include hard float machine description and
9172 modify patterns for hard float configuration.
9173 * config/nds32/nds32.opt: New options for hard float configuration.
9174 * config/nds32/predicates.md: New predicates for hard float
9177 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9179 * common/config/nds32/nds32-common.c
9180 (nds32_option_optimization_table): Enable -mreleax-hint by default.
9182 2018-04-05 Jakub Jelinek <jakub@redhat.com>
9185 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
9186 CONSTRUCTOR_ELT (ctor, ...)->value.
9188 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
9191 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
9193 2018-04-05 Tom de Vries <tom@codesourcery.com>
9196 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
9199 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9200 Kito Cheng <kito.cheng@gmail.com>
9202 * config/nds32/constraints.md (U33): Fine-tune checking condition.
9203 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
9204 * config/nds32/nds32.h (nds32_16bit_address_type): Add
9205 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
9207 2018-04-05 Shiva Chen <shiva0217@gmail.com>
9208 Kito Cheng <kito.cheng@gmail.com>
9210 * config/nds32/constraints.md (Ufe): New memory constraint.
9211 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
9212 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
9213 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
9215 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
9216 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
9218 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9220 * config/nds32/nds32.md: Use optimize_size in the condition for
9221 alu-shift instructions.
9223 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9225 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
9227 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9229 * config/nds32/nds32.md (negsi2): Refine pattern.
9231 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9232 Chung-Ju Wu <jasonwucj@gmail.com>
9234 * config/nds32/iterators.md (shift_rotate): New code iterator.
9235 (shift): New code attribute.
9236 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
9237 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
9238 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
9239 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
9240 bit-wise operations.
9241 (andsi3, *andsi3): Ditto.
9242 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
9243 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
9244 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
9245 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
9246 nds32_ior_operand, nds32_xor_operand): New predicates.
9248 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9250 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
9251 (addsi3, subsi3): ... this.
9253 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9255 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
9257 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9259 * config/nds32/nds32.md: Adjust indention.
9261 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
9263 * config/nds32/nds32.md (feature): New attribute.
9265 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
9267 * config/nds32/nds32.md (subtype): New attribute.
9269 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
9272 * config/arm/arm-builtins.c (arm_expand_builtin): Change
9273 expansion to perform a bitwise AND of the argument followed by a
9274 boolean negation of the result.
9276 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
9278 PR rtl-optimization/84878
9279 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
9280 the basic block. Assert the use reference is not artificial and that
9281 it has an associated insn.
9283 2018-04-04 Michael Matz <matz@suse.de>
9285 * builtins.c (compute_objsize): Pass correct operand
9286 to array_at_struct_end_p.
9288 2018-04-04 Richard Biener <rguenther@suse.de>
9291 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
9292 from contexts for DINFO_LEVEL_TERSE and below.
9294 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9296 * config/nds32/nds32-doubleword.md (move_<mode>): Require
9297 resiter_operand condition.
9298 * config/nds32/nds32.md (*move<mode>): Ditto.
9300 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9301 Monk Chiang <sh.chiang04@gmail.com>
9303 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
9305 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9307 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
9309 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9310 Kito Cheng <kito.cheng@gmail.com>
9312 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
9313 nds32_cond_code_str, output_cond_branch,
9314 output_cond_branch_compare_zero, nds32_expand_cbranch,
9315 nds32_expand_cstore, nds32_expand_movcc,
9316 nds32_output_cbranchsi4_equality_zero,
9317 nds32_output_cbranchsi4_equality_reg,
9318 nds32_output_cbranchsi4_equality_reg_or_const_int,
9319 nds32_output_cbranchsi4_greater_less_zero: New functions.
9320 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
9321 nds32_expand_cstore, nds32_expand_movcc,
9322 nds32_output_cbranchsi4_equality_zero,
9323 nds32_output_cbranchsi4_equality_reg,
9324 nds32_output_cbranchsi4_equality_reg_or_const_int,
9325 nds32_output_cbranchsi4_greater_less_zero): Declare.
9326 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
9327 nds32_rimm11s_operand): New predicates.
9328 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
9329 * config/nds32/nds32.md: Rewrite all the branch and conditional move
9332 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
9334 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
9335 * config/nds32/nds32.md: Ditto.
9336 * config/nds32/pipelines.md: Ditto.
9338 2018-04-04 Richard Biener <rguenther@suse.de>
9340 PR tree-optimization/85168
9341 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
9342 propagating abnormals.
9344 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9346 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
9348 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
9349 Kito Cheng <kito.cheng@gmail.com>
9351 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
9352 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
9353 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
9354 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
9355 * config/nds32/nds32.md (sibcall_internal): New.
9356 (sibcall_register): Remove.
9357 (sibcall_immediate): Remove.
9358 (sibcall_value_internal): New.
9359 (sibcall_value_register): Remove.
9360 (sibcall_value_immediate): Remove.
9361 * config/nds32/predicates.md (nds32_general_register_operand): New.
9362 (nds32_call_address_operand): New.
9364 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9366 PR rtl-optimization/85167
9367 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
9368 bb_defs if *split_p, instead preinitialize it to NULL.
9370 PR tree-optimization/85156
9371 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
9372 evaluating the argument multiple times.
9374 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
9376 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
9378 (_mm_cvtpd_ps): Likewise.
9379 (_mm_cvttpd_epi32): Likewise.
9380 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
9381 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
9382 vector, pixel, and bool following altivec.h include.
9384 2018-04-03 Martin Sebor <msebor@redhat.com>
9386 * doc/extend.texi (Common Function Attributes): Clarify.
9387 (const attribute): Likewise.
9388 (pure attribute): Likewise.
9390 2018-04-03 Jakub Jelinek <jakub@redhat.com>
9393 * config/i386/i386.c (ix86_expand_vector_set): Use
9394 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
9396 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
9398 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
9399 instructions when changing rounding bits to preserve precision bits
9400 in the x87 control word.
9402 2018-04-03 Martin Liska <mliska@suse.cz>
9404 PR tree-optimization/82491
9405 * rtl.h (strip_offset_and_add): Replace += suboffset with
9406 poly_uint64 () + suboffset.
9408 2018-03-29 Martin Liska <mliska@suse.cz>
9409 Martin Jambor <mjambor@suse.cz>
9412 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
9413 param_type is not an integral or pointer type.
9415 2018-04-03 Richard Biener <rguenther@suse.de>
9417 * sese.h (recompute_all_dominators): Remove.
9419 2018-04-02 Martin Sebor <msebor@redhat.com>
9421 * doc/invoke.texi (-Wrestrict): Fix typos.
9423 2018-04-02 Jim Wilson <jimw@sifive.com>
9425 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
9426 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
9427 (<optab>di3, <optab>si3_extend): Likewise.
9428 (<optab>si3_mask, <optab>si3_mask_1): New.
9429 (<optab>di3_mask, <optab>di3_mask_1): New.
9430 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
9431 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
9432 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
9434 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
9436 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
9439 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
9441 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
9442 (nds32_canonicalize_comparison): New function.
9444 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9445 Kito Cheng <kito.cheng@gmail.com>
9446 Kuan-Lin Chen <kuanlinchentw@gmail.com>
9448 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
9449 * config/nds32/constants.md (unspec_volatile_element): Add
9450 UNSPEC_VOLATILE_RELAX_GROUP.
9451 * config/nds32/nds32-relax-opt.c: New file.
9452 * config/nds32/nds32-predicates.c
9453 (nds32_symbol_load_store_p): New function.
9454 * config/nds32/nds32-protos.h
9455 (nds32_symbol_load_store_p): Declare function.
9456 (make_pass_nds32_relax_opt): Declare new rtl pass function.
9457 * config/nds32/nds32.c
9458 (nds32_register_pass): New function to register pass.
9459 (nds32_register_passes): New function to register passes.
9460 * config/nds32/nds32.md (relax_group): New pattern.
9461 * config/nds32/nds32.opt (mrelax-hint): New option.
9462 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
9464 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
9466 * config/nds32/t-nds32: Modify files dependency.
9468 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9470 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
9471 (PROFILE_HOOK): Define its implementation.
9473 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
9475 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
9476 type and 32-bit size.
9478 2018-04-01 Jakub Jelinek <jakub@redhat.com>
9481 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
9482 (V_128_256): New mode iterator.
9483 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
9484 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
9485 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
9487 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
9488 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
9490 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
9493 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
9494 NaN inputs correctly.
9496 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
9499 * config/rs6000/vsx.md (??r): New mode attribute.
9500 (*vsx_mov<mode>_64bit): Use it.
9501 (*vsx_mov<mode>_32bit): Likewise.
9503 2018-03-30 Martin Sebor <msebor@redhat.com>
9505 PR tree-optimization/84818
9506 * builtins.c (check_access): Use warning_n.
9508 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9511 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
9513 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
9516 2018-03-30 Julia Koval <julia.koval@intel.com>
9519 * x86-tune.def (movx, partial_reg_dependency): Enable for
9522 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
9525 * lra-constraints.c (process_alt_operands): Move setting
9526 this_alternative_matches below.
9528 2018-03-29 Martin Liska <mliska@suse.cz>
9531 * doc/invoke.texi: Document how LTO works with debug info.
9532 Describe auto-load support of binutils. Mention 'x86-64'
9533 as valid option value of -march option.
9535 2018-03-29 Jakub Jelinek <jakub@redhat.com>
9537 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
9540 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
9541 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
9542 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
9545 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9548 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
9549 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
9550 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
9551 for RS6000_BTM_POWERPC64.
9552 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
9553 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
9554 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
9559 2018-03-28 Carl Love <cel@us.ibm.com>
9562 2017-09-27 Carl Love <cel@us.ibm.com>
9564 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
9565 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
9566 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
9569 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9571 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
9572 instead of __vector bool.
9573 (_mm_max_pu8): Likewise.
9574 (_mm_min_pi16): Likewise.
9576 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
9579 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
9580 (DIVWEUO): Likewise.
9582 (DIVDEUO): Likewise.
9583 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
9584 DIVWEUO and DIVDEUO.
9585 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
9586 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
9587 (div_extend): Likewise.
9588 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
9590 (__builtin_divweuo): Likewise.
9591 (__builtin_divdeo): Likewise.
9592 (__builtin_divdeuo): Likewise.
9594 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9597 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
9598 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
9600 PR tree-optimization/82004
9601 * gimple-match-head.c (optimize_pow_to_exp): New function.
9602 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
9603 Don't fold to exp if optimize_pow_to_exp is false.
9605 2018-03-28 Martin Liska <mliska@suse.cz>
9608 * calls.c (initialize_argument_information): Fix trailing space.
9609 * common.opt: Fix typo and provide better explanation for
9610 -fsanitize-coverage option.
9611 * config/i386/i386.opt: Fix typo.
9613 2018-03-28 Jakub Jelinek <jakub@redhat.com>
9614 Martin Liska <mliska@suse.cz>
9617 * gimplify.c (asan_poison_variable): Don't do the check for
9618 gimplify_omp_ctxp here.
9619 (gimplify_decl_expr): Do it here.
9620 (gimplify_target_expr): Likewise.
9622 2018-03-28 Martin Liska <mliska@suse.cz>
9625 * config/i386/i386.c (ix86_function_arg_advance): Do not call
9626 chkp_type_bounds_count if MPX is not enabled.
9628 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
9630 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
9632 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
9635 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
9636 function to create the function decl for complex long double
9637 multiply and divide for -mabi=ieeelongdouble.
9638 (init_float128_ieee): Call it.
9640 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
9643 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
9644 -fcf-protection=branch -mibt.
9645 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
9647 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9650 * config/arm/arm.c (arm_valid_symbolic_address): Handle
9651 arm_word_relocations.
9653 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
9656 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
9657 extern array declarations.
9659 2018-03-27 Richard Biener <rguenther@suse.de>
9662 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
9663 explicit single_use checks.
9665 2018-03-27 Richard Biener <rguenther@suse.de>
9667 PR tree-optimization/85082
9668 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
9671 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9673 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
9674 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
9675 Turn on fasynchronous-unwind-tables and funwind-tables.
9677 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
9680 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
9681 (*bmi_blsr_<mode>_ccz): Ditto.
9683 2018-03-26 Tom de Vries <tom@codesourcery.com>
9685 PR tree-optimization/85063
9686 * omp-general.c (offloading_function_p): New function. Factor out
9688 * omp-offload.c (pass_omp_target_link::gate): ... here.
9689 * omp-general.h (offloading_function_p): Declare.
9690 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
9691 with attribute omp declare target for offloading functions.
9693 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
9695 PR tree-optimization/84005
9696 * tree-data-ref.h (get_base_for_alignment): Declare.
9697 * tree-data-ref.c (get_base_for_alignment_1): New function.
9698 (get_base_for_alignment): Likewise.
9699 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
9700 get_base_for_alignment to find a suitable base object, instead
9701 of always using drb->base_address.
9703 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9706 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
9707 known size by default.
9709 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
9712 * lra-constraints.c (process_alt_operands): Don't match BLKmode
9713 and non BLKmode operands.
9715 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9718 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
9719 Clean up attributes.
9721 2018-03-23 Richard Biener <rguenther@suse.de>
9724 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
9725 we are going to emit early debug for LTO.
9727 2018-03-23 Jakub Jelinek <jakub@redhat.com>
9730 * function.c (match_asm_constraints_1): Don't optimize if input
9731 doesn't satisfy general_operand predicate for output's mode.
9734 * alias.c (write_dependence_p): Don't require for x_canonicalized
9735 non-VOIDmode if x has VOIDmode.
9738 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
9739 just don't try to optimize it rather than assert it never happens.
9741 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9743 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
9744 macro expansions for definition of ST_INTERNAL_<mode> and
9745 LD_INTERNAL_<mode> builtins.
9746 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
9748 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
9750 (altivec_expand_st_builtin): Likewise.
9751 (altivec_expand_builtin): Remove calls to deleted functions.
9752 (rs6000_address_for_altivec): Delete this function.
9753 * config/rs6000/vector.md: Remove expands for
9754 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
9756 2018-03-22 Sudakshina Das <sudi.das@arm.com>
9759 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
9760 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
9761 re-computing once computed.
9762 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
9763 (arm_init_machine_status): Initialize
9764 machine->static_chain_stack_bytes.
9766 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9769 * doc/extend.texi: Add four new prototypes for vec_ld.
9770 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
9771 definitions for more logical presentation.
9772 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
9773 entries for V1TI variants of __builtin_altivec_ld builtin.
9774 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
9775 handling of V1TI variant of LVX icode pattern.
9776 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
9777 (rs6000_gimple_fold_builtin): Likewise.
9778 (altivec_init_builtins): Add code to define
9779 __builtin_altivec_lvx_v1ti function.
9781 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9784 * function.c (match_asm_constraints_1): Don't do the optimization
9785 if input isn't a REG, SUBREG, MEM or constant.
9787 2018-03-22 Tom de Vries <tom@codesourcery.com>
9789 PR tree-optimization/84956
9790 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
9791 bb_has_abnormal_pred.
9793 2018-03-22 Jakub Jelinek <jakub@redhat.com>
9796 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
9797 DECL_INITIAL (decl) to decl at the end.
9798 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
9801 2018-03-21 Joseph Myers <joseph@codesourcery.com>
9803 * doc/extend.texi (__builtin_tgmath): Document when complex
9804 integer types are treated as _Complex _Float64.
9806 2018-03-21 Tom de Vries <tom@codesourcery.com>
9808 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
9810 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9812 PR tree-optimization/84960
9813 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
9814 if it is ENTRY block, move them into single succ of ENTRY in that case.
9816 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
9818 PR tree-optimization/84811
9819 * poly-int.h (poly_span_traits): Remove the T3 parameter and
9820 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
9821 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
9822 (known_subrange_p): Update accordingly. Cast each value involved
9823 in the size comparison, rather than casting the result of the
9826 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9828 PR tree-optimization/84982
9829 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
9830 by flipping the least significant bit rather than all bits from
9831 bitpos to bitpos + bitsize - 1.
9833 2018-03-21 Nathan Sidwell <nathan@acm.org>
9835 * doc/extend.texi (Deprecated Features): Remove mention of
9836 long-deleted deprecations.
9838 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9841 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
9842 * configure: Regenerate.
9844 2018-03-21 Tom de Vries <tom@codesourcery.com>
9846 PR tree-optimization/83126
9847 * tree-parloops.c (num_phis): New function.
9848 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
9850 2018-03-21 Nathan Sidwell <nathan@acm.org>
9852 * doc/extend.texi (Deprecated Features): Update deprecated flags,
9853 mention anon-struct/union members and trailing attributes.
9855 2018-03-21 Bin Cheng <bin.cheng@arm.com>
9857 PR tree-optimization/84969
9858 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
9859 builtin memset partitions if they set different rhs values.
9861 2018-03-21 Jakub Jelinek <jakub@redhat.com>
9863 PR rtl-optimization/84989
9864 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
9865 VEC_DUPLICATE with scalar result mode.
9867 2018-03-21 Martin Liska <mliska@suse.cz>
9870 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
9871 not intended return statement.
9873 2018-03-21 Martin Liska <mliska@suse.cz>
9876 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
9877 (chkp_find_bound_slots_1): Limit number of iterations.
9879 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
9882 * Minor grammar fixes for x86 options.
9884 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9887 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
9888 holding REG_CFA_RESTORE notes, instead turn them into a USE.
9890 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
9893 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
9894 (altivec_lvx_<mode>_1op): Likewise.
9895 (altivec_stvx_<mode>_2op): Likewise.
9896 (altivec_stvx_<mode>_1op): Likewise.
9897 (altivec_lvx_<VM2:mode>): New define_expand.
9898 (altivec_stvx_<VM2:mode>): Likewise.
9899 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
9900 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9901 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
9902 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
9903 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
9904 (rs6000_gen_lvx): Likewise.
9905 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
9906 (altivec_expand_stv_builtin): Likewise.
9907 (altivec_expand_builtin): Likewise.
9908 * config/rs6000/vector.md: Likewise.
9910 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9913 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
9916 2018-03-20 Richard Biener <rguenther@suse.de>
9919 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
9920 sign-conversions as zero, fall back to standard scalar_stmt
9923 2018-03-20 Martin Liska <mliska@suse.cz>
9926 * predict.c (rebuild_frequencies): Handle case when we have
9927 PROFILE_ABSENT, but flag_guess_branch_prob is false.
9929 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9932 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
9933 flag_section_anchors.
9934 * varasm.c (use_blocks_for_decl_p): Remove hack for
9935 dw2_force_const_mem.
9938 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
9940 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
9941 be created, use lowpart_subreg of operands[0] rather than operands[0]
9943 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
9944 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
9945 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
9946 and n constraint instead of aarch64_shift_imm_di and Usd.
9947 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
9948 (*aarch64_<optab>_reg_minus<mode>3): ... this.
9950 2018-03-20 Sudakshina Das <sudi.das@arm.com>
9953 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
9954 to favor GPR over NEON registers.
9955 (<shift>di3_neon): Likewise.
9957 2018-03-20 Tom de Vries <tom@codesourcery.com>
9960 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
9961 (nvptx_process_pars): Emit bar.sync asap and alap.
9963 2018-03-20 Tom de Vries <tom@codesourcery.com>
9966 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
9967 seen_label if seen_label is already set.
9969 2018-03-20 Jakub Jelinek <jakub@redhat.com>
9972 * config/i386/i386.c (fold_builtin_cpu): For features above 31
9973 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
9974 Use 1U instead of 1. Formatting fixes.
9977 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
9978 instead of TREE_TYPE (s1) for the return value.
9980 2018-03-19 Jakub Jelinek <jakub@redhat.com>
9982 PR tree-optimization/84946
9983 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
9984 bitsize + bitsize in poly_uint64 rather than poly_int64.
9987 * dwarf2asm.c: Include fold-const.c.
9988 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
9989 of decl rather than decl itself.
9991 PR rtl-optimization/84643
9992 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
9994 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
9997 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
9998 calling assemble_variable.
10000 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10003 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10004 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10006 2018-03-19 Jim Wilson <jimw@sifive.com>
10009 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10010 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10011 (riscv_first_stack_step): Likewise.
10012 (riscv_option_override): Use STACK_BOUNDARY instead of
10013 MIN_STACK_BOUNDARY.
10014 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10015 MIN_STACK_BOUNDARY.
10016 (BIGGEST_ALIGNMENT): Set to 128.
10017 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10018 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10021 2018-03-19 Richard Biener <rguenther@suse.de>
10023 PR tree-optimization/84933
10024 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10025 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10027 2018-03-19 Richard Biener <rguenther@suse.de>
10029 PR tree-optimization/84859
10030 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10031 (cond_if_else_store_replacement): Perform sinking operation on
10032 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10033 Generalize what a BB with a single eligible store is.
10035 2018-03-19 Richard Biener <rguenther@suse.de>
10037 PR tree-optimization/84929
10038 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10039 chrec_is_positive against non-chrec arg.
10041 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10044 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10046 2018-03-18 Martin Liska <mliska@suse.cz>
10048 PR rtl-optimization/84635
10049 * regrename.c (build_def_use): Use matches_mode only when
10052 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10054 PR tree-optimization/84913
10055 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10056 vectorize chains of COND_EXPRs.
10058 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10060 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10062 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10064 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
10066 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10068 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
10070 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10071 Kito Cheng <kito.cheng@gmail.com>
10073 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
10074 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
10075 (nds32_adjust_reg_alloc_order): New function.
10076 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
10078 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
10080 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
10081 nds32_print_operand, nds32_print_operand_address): Use
10082 HOST_WIDE_INT_PRINT_DEC instead.
10084 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
10086 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
10088 2018-03-17 Jakub Jelinek <jakub@redhat.com>
10091 * config/i386/i386.c (initial_ix86_tune_features,
10092 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
10093 unsigned long long.
10094 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
10095 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
10096 rather than 1u << ix86_tune. Formatting fix.
10097 (ix86_option_override_internal): Change ix86_arch_mask from
10098 unsigned int to unsigned HOST_WIDE_INT, initialize to
10099 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
10100 (ix86_function_specific_restore): Likewise.
10102 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10105 * postreload.c (reload_combine_recognize_pattern): Perform
10106 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
10107 truncate_int_for_mode the result for the destination's mode.
10110 * hsa-gen.c (mem_type_for_type): Fix comment typo.
10111 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
10113 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10116 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
10119 * lra-assigns.c (lra_split_hard_reg_for): Don't use
10120 regno_allocno_class_array and sorted_pseudos.
10121 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
10122 insns where regno is used.
10124 2018-03-16 Martin Liska <mliska@suse.cz>
10127 * multiple_target.c (create_dispatcher_calls): Redirect
10128 reference in the symbol table.
10130 2018-03-16 Martin Liska <mliska@suse.cz>
10133 * multiple_target.c (create_dispatcher_calls): Redirect also
10136 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10140 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
10141 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
10142 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
10144 2018-03-16 Julia Koval <julia.koval@intel.com>
10146 * doc/invoke.texi (Skylake Server): Add CLWB.
10147 Cannonlake): Remove CLWB.
10149 2018-03-16 Jakub Jelinek <jakub@redhat.com>
10151 PR tree-optimization/84841
10152 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
10154 (FLOAT_ONE_CONST_TYPE): Define.
10155 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
10156 (sort_by_operand_rank): Put entries with higher constant_type last
10157 rather than first to match comments.
10159 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
10161 * config/nios2/nios2.md (movsi_internal): Fix thinko in
10164 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10167 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
10168 check and use address of target always.
10170 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
10173 * config/i386/i386.c (indirect_thunk_needed): Update comments.
10174 (indirect_thunk_bnd_needed): Likewise.
10175 (indirect_thunks_used): Likewise.
10176 (indirect_thunks_bnd_used): Likewise.
10177 (indirect_return_needed): New.
10178 (indirect_return_bnd_needed): Likewise.
10179 (output_indirect_thunk_function): Add a bool argument for
10181 (output_indirect_thunk_function): Don't generate alias for
10182 function return thunk.
10183 (ix86_code_end): Call output_indirect_thunk_function to generate
10184 function return thunks.
10185 (ix86_output_function_return): Set indirect_return_bnd_needed
10186 and indirect_return_needed instead of indirect_thunk_bnd_needed
10187 and indirect_thunk_needed.
10189 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
10191 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
10192 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
10193 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
10195 2018-03-15 David Malcolm <dmalcolm@redhat.com>
10196 Paul Hua <paul.hua.gm@gmail.com>
10199 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
10201 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
10203 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
10204 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
10205 resp. SFmode cases.
10207 2018-03-15 Tamar Christina <tamar.christina@arm.com>
10210 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10211 instead of GET_MODE_SIZE when comparing Units.
10213 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10216 * varasm.c (hash_section): Return an unchangeble hash value
10217 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
10218 Return !aarch64_can_use_per_function_literal_pools_p ().
10220 2018-03-15 Jakub Jelinek <jakub@redhat.com>
10223 * optabs.c (emit_conditional_move): Pass address of cmode's copy
10224 rather than address of cmode as last argument to prepare_cmp_insn.
10226 2018-03-15 Julia Koval <julia.koval@intel.com>
10228 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
10229 F_AVX512VNNI, F_AVX512BITALG): New.
10231 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
10234 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
10235 insn for floating-point loads and stores.
10237 2018-03-14 Carl Love <cel@us.ibm.com>
10239 * config/rs6000/rs6000-c.c: Add macro definitions for
10240 ALTIVEC_BUILTIN_VEC_PERMXOR.
10241 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
10242 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
10243 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
10244 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
10246 * config/doc/extend.texi: Add prototypes for vec_permxor.
10248 2018-03-14 David Malcolm <dmalcolm@redhat.com>
10251 * diagnostic-show-locus.c (class layout_point): Convert m_line
10252 from int to linenum_type.
10253 (line_span::comparator): Use linenum "compare" function when
10254 comparing line numbers.
10255 (test_line_span): New function.
10256 (layout_range::contains_point): Convert param "row" from int to
10258 (layout_range::intersects_line_p): Likewise.
10259 (layout::will_show_line_p): Likewise.
10260 (layout::print_source_line): Likewise.
10261 (layout::should_print_annotation_line_p): Likewise.
10262 (layout::print_annotation_line): Likewise.
10263 (layout::print_leading_fixits): Likewise.
10264 (layout::annotation_line_showed_range_p): Likewise.
10265 (struct line_corrections): Likewise for field m_row.
10266 (line_corrections::line_corrections): Likewise for param "row".
10267 (layout::print_trailing_fixits): Likewise.
10268 (layout::get_state_at_point): Likewise.
10269 (layout::get_x_bound_for_row): Likewise.
10270 (layout::print_line): Likewise.
10271 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
10272 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
10273 * input.c (selftest::test_linenum_comparisons): New function.
10274 (selftest::input_c_tests): Call it.
10275 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
10276 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
10277 * selftest.h (ASSERT_GT): New macro.
10278 (ASSERT_GT_AT): New macro.
10279 (ASSERT_LT): New macro.
10280 (ASSERT_LT_AT): New macro.
10282 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
10284 PR rtl-optimization/84780
10285 * combine.c (distribute_links): Don't make a link based on pc_rtx.
10287 2018-03-14 Martin Liska <mliska@suse.cz>
10289 * tree.c (record_node_allocation_statistics): Use
10290 get_stats_node_kind.
10291 (get_stats_node_kind): New function extracted from
10292 record_node_allocation_statistics.
10293 (free_node): Use get_stats_node_kind.
10295 2018-03-14 Richard Biener <rguenther@suse.de>
10297 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
10298 that the value-set of ANTIC_IN doesn't grow.
10301 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
10303 (BB_VISITED_WITH_VISITED_SUCCS): New define.
10304 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
10306 2018-03-14 Julia Koval <julia.koval@intel.com>
10308 * config.gcc (icelake-client, icelake-server): New.
10310 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
10311 (initial_ix86_arch_features): Ditto.
10312 (PTA_SKYLAKE): Add SGX.
10313 (PTA_ICELAKE): Remove.
10314 (PTA_ICELAKE_CLIENT): New.
10315 (PTA_ICELAKE_SERVER): New.
10316 (ix86_option_override_internal): Split up icelake on icelake client and
10318 (get_builtin_code_for_version): Ditto.
10319 (fold_builtin_cpu): Ditto.
10320 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
10321 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
10322 * config/i386/i386.h (processor_type): Ditto.
10323 * doc/invoke.texi: Ditto.
10325 2018-03-14 Jakub Jelinek <jakub@redhat.com>
10328 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
10329 INTEGER_CST offset, add it together with bitpos / 8 and
10330 sign extend based on POINTER_SIZE.
10334 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
10337 * config/i386/constraints.md (Yc): New register constraint.
10338 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
10339 Use Yc constraint for alternative 2 of operand 0. Remove
10340 preferred_for_speed attribute.
10342 2018-03-14 Richard Biener <rguenther@suse.de>
10344 PR tree-optimization/84830
10345 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
10346 with the old one to avoid oscillations.
10348 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
10351 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
10353 (assign_by_spills): Return a flag of reload assignment failure.
10354 Do not process the reload assignment failures. Do not spill other
10355 reload pseudos if they has the same reg class. Update n if
10357 (lra_assign): Add a return arg. Set up from the result of
10358 assign_by_spills call.
10359 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10360 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10361 usage_insns if it is not NULL.
10362 (spill_hard_reg_in_range): New function.
10363 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10364 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10365 function prototypes.
10366 (lra_assign): Change prototype.
10367 * lra.c (lra): Add code to deal with fails by splitting hard reg
10370 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
10372 * config/riscv/riscv.opt (mrelax): New option.
10373 * config/riscv/riscv.c (riscv_file_start): Emit ".option
10374 "norelax" when riscv_mrelax is disabled.
10375 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
10377 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10380 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
10381 reassociation for int modes.
10383 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10385 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
10386 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
10388 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
10389 * config/aarch64/aarch64-sve.md
10390 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
10391 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
10392 (*extend<mode><Vwide>2): Rename to...
10393 (aarch64_sve_extend<mode><Vwide>2): ...this.
10394 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
10395 renaming the old pattern to...
10396 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
10398 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
10399 define_expand, renaming the old pattern to...
10400 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
10401 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
10402 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
10403 account when deciding which SVE instruction the optab should use.
10404 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
10406 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10408 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
10409 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
10410 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
10411 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
10412 (tlsdesc_small_<mode>): Turn a define_expand and use
10413 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
10414 (tlsdesc_small_advsimd_<mode>): ...this.
10415 (tlsdesc_small_sve_<mode>): New pattern.
10417 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
10419 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
10420 (UNSPEC_UMUL_HIGHPART): New constants.
10421 (MUL_HIGHPART): New int iteraor.
10422 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
10423 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
10425 (*<su>mul<mode>3_highpart): New define_insn.
10427 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10430 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10433 2018-03-13 Martin Liska <mliska@suse.cz>
10436 * (sem_item_optimizer::sem_item_optimizer): Initialize new
10438 (sem_item_optimizer::~sem_item_optimizer): Release it.
10439 (sem_item_optimizer::merge_classes): Register variable aliases.
10440 (sem_item_optimizer::fixup_pt_set): New function.
10441 (sem_item_optimizer::fixup_points_to_sets): Likewise.
10442 * ipa-icf.h: Declare new variables and functions.
10444 2018-03-13 Jakub Jelinek <jakub@redhat.com>
10446 PR middle-end/84834
10447 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
10448 integer_pow2p@2 and test integer_pow2p in condition.
10449 (A < 0 ? C : 0): Similarly for @1.
10451 PR middle-end/84831
10452 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
10453 characters starting at p contain '\0' character, don't look beyond
10457 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
10458 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
10461 * reg-stack.c (change_stack): Change update_end var from int to
10462 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
10463 also call set_block_for_insn on the newly added insns and rescan.
10466 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
10467 on the last operand.
10470 * tree.c (stabilize_reference_1): Return save_expr (e) for
10471 STATEMENT_LIST even if it doesn't have side-effects.
10473 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
10475 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
10477 2018-03-12 Renlin Li <renlin.li@arm.com>
10479 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
10480 aarch64_output_scalar_simd_mov_immediate.
10482 2018-03-12 Martin Sebor <msebor@redhat.com>
10484 PR tree-optimization/83456
10485 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
10486 for perfectly overlapping calls to memcpy.
10487 (gimple_fold_builtin_memory_chk): Same.
10488 (gimple_fold_builtin_strcpy): Handle no-warning.
10489 (gimple_fold_builtin_stxcpy_chk): Same.
10490 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
10492 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10494 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
10495 parameter. Use it for SFmode.
10496 (rs6000_function_arg_advance_1): Adjust.
10497 (rs6000_function_arg): Adjust.
10498 (rs6000_gimplify_va_arg): Pass false for that new parameter.
10500 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
10502 PR rtl-optimization/84169
10503 PR rtl-optimization/84780
10504 * combine.c (can_combine_p): Check for a 2-insn combination whether
10505 the destination register is used between the two insns, too.
10507 2018-03-12 Richard Biener <rguenther@suse.de>
10509 PR tree-optimization/84803
10510 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
10511 for refs DR analysis didn't process.
10513 2018-03-12 Richard Biener <rguenther@suse.de>
10515 PR tree-optimization/84777
10516 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
10517 force-vectorize loops ignore whether we are optimizing for size.
10519 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
10521 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
10522 (TARGET_MD_ASM_ADJUST): Define.
10524 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
10525 Kito Cheng <kito.cheng@gmail.com>
10526 Chung-Ju Wu <jasonwucj@gmail.com>
10528 * config/nds32/nds32.c (nds32_compute_stack_frame,
10529 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
10530 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
10531 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
10532 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
10533 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
10534 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
10535 * config/nds32/nds32.md (prologue, epilogue): Use macro
10536 NDS32_V3PUSH_AVAILABLE_P to do checking.
10538 2018-03-11 Jakub Jelinek <jakub@redhat.com>
10541 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
10542 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
10543 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
10544 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
10545 addition of most attributes on !orig_type_die or the attribute not
10546 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
10548 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10549 Chung-Ju Wu <jasonwucj@gmail.com>
10551 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
10552 __NDS32_VH__ macro.
10553 * config/nds32/nds32.opt (mvh): New option.
10555 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10556 Chung-Ju Wu <jasonwucj@gmail.com>
10558 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
10560 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
10561 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
10564 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
10565 Chung-Ju Wu <jasonwucj@gmail.com>
10567 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
10569 * config/nds32/nds32-multiple.md (strlensi): New pattern.
10570 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
10572 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
10573 Kito Cheng <kito.cheng@gmail.com>
10574 Chung-Ju Wu <jasonwucj@gmail.com>
10576 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
10577 UNSPEC_FFMISM and UNSPEC_FLMISM.
10578 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
10579 for ffb, ffmism and flmism.
10580 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
10581 (unspec_ffmism): Ditto.
10582 (unspec_flmism): Ditto.
10583 (nds32_expand_builtin_impl): Check if string extension is available.
10584 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
10585 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
10587 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
10590 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10593 * lra-assigns.c (assign_by_spills): Return a flag of reload
10594 assignment failure. Do not process the reload assignment
10595 failures. Do not spill other reload pseudos if they has the same
10597 (lra_assign): Add a return arg. Set up from the result of
10598 assign_by_spills call.
10599 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10600 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10601 usage_insns if it is not NULL.
10602 (spill_hard_reg_in_range): New function.
10603 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10604 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10605 function prototypes.
10606 (lra_assign): Change prototype.
10607 * lra.c (lra): Add code to deal with fails by splitting hard reg
10610 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
10613 * config/i386/i386.opt: Replace Enforcment with Enforcement.
10615 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
10618 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
10619 (dw_val_node): Add val_symbolic_view.
10620 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
10621 (symview_upper_bound): New.
10622 (new_line_info_table): Initialize symviews_since_reset.
10623 (dwarf2out_source_line): Count symviews_since_reset and set
10624 symview_upper_bound.
10625 (dw_val_equal_p): Handle symview.
10626 (add_AT_symview): New.
10627 (print_dw_val): Handle symview.
10628 (attr_checksum, attr_checksum_ordered): Likewise.
10629 (same_dw_val_p, size_of_die): Likewise.
10630 (value_format, output_die): Likewise.
10631 (add_high_low_attributes): Use add_AT_symview for entry_view.
10632 (dwarf2out_finish): Reset symview_upper_bound, clear
10635 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
10638 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
10639 Add strict argument and use it.
10640 (rs6000_split_multireg_move): Update for new strict argument.
10641 (mem_operand_gpr): Disallow all non-offsettable addresses.
10642 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
10644 2018-03-09 Jakub Jelinek <jakub@redhat.com>
10647 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
10648 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
10649 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
10652 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
10653 decl, use remap_type if we want to use the type.
10655 2018-03-09 Martin Sebor <msebor@redhat.com>
10657 PR tree-optimization/84526
10658 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
10660 (builtin_access::generic_overlap): Be prepared to handle non-array
10663 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
10665 PR rtl-optimization/84682
10666 * lra-constraints.c (process_address_1): Check is_address flag
10667 for address constraints.
10668 (process_alt_operands): Likewise.
10669 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
10670 preprocess_constraints.
10671 * recog.h (preprocess_constraints): Add oploc parameter.
10673 * recog.c (preprocess_constraints): Test address_operand for
10674 CT_ADDRESS constraints.
10676 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
10679 * lra-assigns.c (assign_by_spills): Return a flag of reload
10680 assignment failure. Do not process the reload assignment
10681 failures. Do not spill other reload pseudos if they has the same
10683 (lra_assign): Add a return arg. Set up from the result of
10684 assign_by_spills call.
10685 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
10686 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
10687 usage_insns if it is not NULL.
10688 (spill_hard_reg_in_range): New function.
10689 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
10690 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
10691 function prototypes.
10692 (lra_assign): Change prototype.
10693 * lra.c (lra): Add code to deal with fails by splitting hard reg
10696 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10699 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
10700 Accept complain bool parameter. Only emit errors if it is true.
10701 (arm_parse_cpu_option_name): Likewise.
10702 (arm_target_thumb_only): Adjust callers of the above.
10703 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
10704 prototype to take a default true bool parameter.
10705 (arm_parse_arch_option_name): Likewise.
10707 2018-03-09 David Malcolm <dmalcolm@redhat.com>
10708 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
10712 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
10713 * configure: Regenerate.
10714 * configure.ac ("linker --version-script option"): New.
10715 ("linker soname option"): New.
10717 2018-03-09 Richard Biener <rguenther@suse.de>
10719 PR tree-optimization/84775
10720 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
10721 immediate uses of predicate stmts and mark them modified.
10724 PR tree-optimization/84178
10725 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10727 (version_loop_for_if_conversion): Delay update_ssa call.
10728 (tree_if_conversion): Delay update_ssa until after predicate
10731 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
10734 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
10735 when the function accesses prior frames.
10737 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10740 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
10741 gen_llsym, otherwise call maybe_gen_llsym.
10743 PR inline-asm/84742
10744 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
10745 has ',' character inside of it.
10747 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10750 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
10751 as clobbering CC_REGNUM.
10753 2018-03-08 Richard Biener <rguenther@suse.de>
10755 PR middle-end/84552
10756 * tree-scalar-evolution.c: Include tree-into-ssa.h.
10757 (follow_copies_to_constant): Do not follow SSA names registered
10760 2018-03-08 Richard Biener <rguenther@suse.de>
10762 PR tree-optimization/84178
10763 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
10765 (version_loop_for_if_conversion): Delay update_ssa call.
10766 (tree_if_conversion): Delay update_ssa until after predicate
10769 2018-03-08 David Malcolm <dmalcolm@redhat.com>
10771 PR tree-optimization/84178
10772 * tree-if-conv.c (release_bb_predicate): Remove the
10773 the assertion that the stmts have NULL use_ops.
10774 Discard the statements, asserting that they haven't
10775 yet been added to a BB.
10777 2018-03-08 Richard Biener <rguenther@suse.de>
10779 PR tree-optimization/84746
10780 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
10781 (phi_translate): Pass in destination ANTIC_OUT set.
10782 (phi_translate_1): Likewise. For a simplified result lookup
10783 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
10784 (phi_translate_set): Adjust.
10785 (do_pre_regular_insertion): Likewise.
10786 (do_pre_partial_partial_insertion): Likewise.
10788 2018-03-08 Martin Liska <mliska@suse.cz>
10790 PR gcov-profile/84735
10791 * doc/gcov.texi: Document usage of profile files.
10792 * gcov-io.h: Document changes in the format.
10794 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
10798 * dwarf2out.c (struct dw_line_info_table): Update comments for
10800 (FORCE_RESET_NEXT_VIEW): New.
10801 (FORCE_RESETTING_VIEW_P): New.
10802 (RESETTING_VIEW_P): Check for -1 too.
10803 (ZERO_VIEW_P): Likewise.
10804 (new_line_info_table): Force-reset next view.
10805 (dwarf2out_begin_function): Likewise.
10806 (dwarf2out_source_line): Simplify zero_view_p initialization.
10807 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
10808 view directly. Omit view when omitting .loc at line 0.
10810 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10812 PR tree-optimization/84740
10813 * tree-switch-conversion.c (process_switch): Call build_constructors
10814 only if info.phi_count is non-zero.
10816 PR tree-optimization/84739
10817 * tree-tailcall.c (find_tail_calls): Check call arguments against
10818 DECL_ARGUMENTS (current_function_decl) rather than
10819 DECL_ARGUMENTS (func) when checking for tail recursion.
10821 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10823 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
10824 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
10825 Volker Reichelt's entry and add entries for people that perform
10826 GCC fuzzy testing and report numerous bugs.
10828 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
10831 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
10832 readonly data in sdata, if that is disabled.
10833 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
10834 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
10835 -mreadonly-in-sdata option.
10837 2018-03-07 Martin Sebor <msebor@redhat.com>
10839 PR tree-optimization/84468
10840 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
10841 basic block when looking for nul assignment.
10843 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
10846 * except.h (output_function_exception_table): Adjust prototype.
10847 * except.c (output_function_exception_table): Remove FNNAME parameter
10848 and add SECTION parameter. Ouput one part of the table at a time.
10849 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
10850 the first part of the exception table and emit unwind directives.
10851 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
10852 (i386_pe_seh_cold_init): Likewise.
10853 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
10854 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
10855 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
10856 (ix86_output_call_insn): Emit a nop in one more case for SEH.
10857 * config/i386/winnt.c: Include except.h.
10858 (struct seh_frame_state): Add reg_offset, after_prologue and
10859 in_cold_section fields.
10860 (i386_pe_seh_end_prologue): Set seh->after_prologue.
10861 (i386_pe_seh_cold_init): New function.
10862 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
10863 to seh->in_cold_section.
10864 (seh_emit_push): Record the offset of the push.
10865 (seh_emit_save): Record the offet of the save.
10866 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
10867 Test seh->after_prologue to disregard the epilogue.
10868 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
10869 (i386_pe_end_cold_function): New function.
10871 2018-03-07 Jakub Jelinek <jakub@redhat.com>
10874 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
10875 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
10878 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
10880 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
10881 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10883 PR middle-end/84723
10884 * multiple_target.c: Include tree-inline.h and intl.h.
10885 (expand_target_clones): Diagnose and fail if node->definition and
10886 !tree_versionable_function_p (node->decl).
10888 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
10890 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
10892 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
10893 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
10894 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
10896 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10899 * combine.c (try_combine): Use reg_or_subregno instead of handling
10900 just paradoxical SUBREGs and REGs.
10902 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
10904 * config/arc/arc.c (arc_finalize_pic): Remove function.
10905 (arc_must_save_register): We use single base PIC register, remove
10906 checks to save/restore the PIC register.
10907 (arc_expand_prologue): Likewise.
10908 * config/arc/arc-protos.h (arc_set_default_type_attributes):
10910 (arc_verify_short): Likewise.
10911 (arc_attr_type): Likewise.
10912 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
10913 (walk_stores): Likewise.
10914 (arc_address_cost): Make it static.
10915 (arc_verify_short): Likewise.
10916 (branch_dest): Likewise.
10917 (arc_attr_type): Likewise.
10918 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
10919 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
10920 (arc_final_prescan_insn): Remove inserting the nops due to
10921 hardware hazards. It is done in reorg step.
10922 (insn_length_variant_t): Remove.
10923 (insn_length_parameters_t): Likewise.
10924 (arc_insn_length_parameters): Likewise.
10925 (arc_get_insn_variants): Likewise.
10926 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
10928 2018-03-06 Jakub Jelinek <jakub@redhat.com>
10930 PR inline-asm/84683
10931 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
10934 PR tree-optimization/84687
10935 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10937 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
10939 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10941 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
10942 Rename to ppc_speculation_barrier.
10943 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
10944 __builtin_ppc_speculation_barrier.
10946 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10949 * combine.c (combine_simplify_rtx): Don't try to simplify if
10950 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
10953 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
10955 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
10956 to 32 bytes when compiling for POWER9.
10958 2018-03-05 Jakub Jelinek <jakub@redhat.com>
10961 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
10962 regparm >= 3 with no arg reg available also for calls with
10963 flag_force_indirect_call. Pass decl to ix86_function_regparm.
10966 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
10968 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
10970 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
10973 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
10975 2018-03-05 Richard Biener <rguenther@suse.de>
10977 PR tree-optimization/84486
10978 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
10979 When inserting a __builtin_assume_aligned call set the LHS
10980 SSA name alignment info accordingly.
10982 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
10984 PR tree-optimization/84114
10985 * config/aarch64/aarch64.c (aarch64_reassociation_width)
10986 Avoid reassociation of FLOAT_MODE addition.
10988 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
10990 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
10991 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
10992 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
10993 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
10994 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
10995 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
10996 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
10998 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
10999 __builtin_ia32_wbinvd): New builtins.
11000 (SPECIAL_ARGS2): New.
11001 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11002 (SPECIAL_ARGS2): New.
11003 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11004 (ix86_valid_target_attribute_inner_p): Ditto.
11005 (ix86_init_mmx_sse_builtins): Add special_args2.
11006 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11007 TARGET_WBNOINVD_P): New.
11008 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11009 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11010 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11011 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11012 * config/i386/pconfigintrin.h: New file.
11013 * config/i386/wbnoinvdintrin.h: Ditto.
11014 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11016 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11018 2018-03-05 Richard Biener <rguenther@suse.de>
11020 PR tree-optimization/84670
11021 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11023 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11024 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11025 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11026 doesn't grow if all successors (recursively) were visited at least
11029 2018-03-05 Richard Biener <rguenther@suse.de>
11031 PR tree-optimization/84650
11032 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11033 if executed in the loop pipeline.
11035 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11037 * doc/configfiles.texi (Configuration Files): Move info about
11038 conditionalizing $target-protos.h to...
11039 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11040 differs from $target-protos.h.
11042 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11043 Chung-Ju Wu <jasonwucj@gmail.com>
11045 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11046 * config/nds32/nds32-multiple.md (setmemsi): Define.
11047 * config/nds32/nds32-memory-manipulation.c
11048 (nds32_gen_dup_4_byte_to_word_value): New.
11049 (emit_setmem_word_loop): New.
11050 (emit_setmem_byte_loop): New.
11051 (nds32_expand_setmem_loop): New.
11052 (nds32_expand_setmem_loop_v3m): New.
11053 (nds32_expand_setmem_unroll): New.
11054 (nds32_expand_setmem): New.
11056 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11057 Chung-Ju Wu <jasonwucj@gmail.com>
11059 * config/nds32/nds32-memory-manipulation.c
11060 (nds32_emit_load_store): New.
11061 (nds32_emit_post_inc_load_store): New.
11062 (nds32_emit_mem_move): New.
11063 (nds32_emit_mem_move_block): New.
11064 (nds32_expand_movmemsi_loop_unknown_size): New.
11065 (nds32_expand_movmemsi_loop_known_size): New.
11066 (nds32_expand_movmemsi_loop): New.
11067 (nds32_expand_movmemsi_unroll): New.
11068 (nds32_expand_movmemqi): Rename ...
11069 (nds32_expand_movmemsi): ... to this.
11070 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
11071 (movmemsi): ... to this.
11072 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
11073 (nds32_expand_movmemsi): ... to this.
11075 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11076 Monk Chiang <sh.chiang04@gmail.com>
11077 Chung-Ju Wu <jasonwucj@gmail.com>
11079 * config/nds32/nds32-protos.h
11080 (nds32_expand_load_multiple): New arguments.
11081 (nds32_expand_store_multiple): Ditto.
11082 (nds32_valid_multiple_load_store): Rename ...
11083 (nds32_valid_multiple_load_store_p): ... to this.
11084 * config/nds32/nds32-memory-manipulation.c
11085 (nds32_expand_load_multiple): Refine implementation.
11086 (nds32_expand_store_multiple): Ditto.
11087 * config/nds32/nds32-multiple.md
11088 (load_multiple): Update nds32_expand_load_multiple interface.
11089 (store_multiple): Update nds32_expand_store_multiple interface.
11090 * config/nds32/nds32-predicates.c
11091 (nds32_valid_multiple_load_store): Rename ...
11092 (nds32_valid_multiple_load_store_p): ... to this and refine
11094 * config/nds32/predicates.md
11095 (nds32_load_multiple_and_update_address_operation): New predicate.
11096 (nds32_store_multiple_and_update_address_operation): New predicate.
11098 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11099 Chung-Ju Wu <jasonwucj@gmail.com>
11101 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
11102 (combo): New attribute.
11103 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
11105 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
11107 * config/nds32/nds32.opt: Change -mcmodel= default value.
11109 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
11110 Monk Chiang <sh.chiang04@gmail.com>
11111 Chung-Ju Wu <jasonwucj@gmail.com>
11113 * config/nds32/constants.md (unspec_element): New enum.
11114 * config/nds32/constraints.md (Umw): New constraint.
11115 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
11116 * config/nds32/nds32-intrinsic.md: Likewise.
11117 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
11118 (nds32_valid_smw_lwm_base_p): New.
11119 (nds32_output_smw_single_word): New.
11120 (nds32_output_lmw_single_word): New.
11121 (nds32_expand_unaligned_load): New.
11122 (nds32_expand_unaligned_store): New.
11123 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
11124 (nds32_output_smw_single_word): Declare.
11125 (nds32_output_lmw_single_word): Declare.
11126 (nds32_expand_unaligned_load): Declare.
11127 (nds32_expand_unaligned_store): Declare.
11128 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
11129 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
11130 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
11131 NDS32_BUILTIN_UASTORE_DW.
11132 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
11135 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
11136 Kito Cheng <kito.cheng@gmail.com>
11137 Chung-Ju Wu <jasonwucj@gmail.com>
11139 * config/nds32/nds32-intrinsic.c
11140 (nds32_expand_builtin_null_ftype_reg): Delete.
11141 (nds32_expand_builtin_reg_ftype_imm): Ditto.
11142 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
11143 (nds32_read_argument): New.
11144 (nds32_legitimize_target): Ditto.
11145 (nds32_legitimize_argument): Ditto.
11146 (nds32_check_constant_argument): Ditto.
11147 (nds32_expand_unop_builtin): Ditto.
11148 (nds32_expand_unopimm_builtin): Ditto.
11149 (nds32_expand_binop_builtin): Ditto.
11150 (nds32_builtin_decl_impl): Ditto.
11151 (builtin_description): Ditto.
11152 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
11153 (nds32_init_builtins_impl): Ditto.
11154 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
11155 (nds32_builtin_decl): New.
11156 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
11157 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
11159 2018-03-02 Jeff Law <law@redhat.com>
11161 * reorg.c (stop_search_p): Handle DEBUG_INSN.
11162 (redundant_insn, fill_simple_delay_slots): Likewise.
11163 (fill_slots_from_thread): Likewise.
11164 * resource.c (mark_referenced_resources): Likewise.
11165 (mark_set_resources, find_dead_or_set_registers): Likewise.
11167 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11169 * substring-locations.h (format_warning_va): Formatting fix for
11170 ATTRIBUTE_GCC_DIAG.
11171 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
11173 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
11174 * substring-locations.c: Include intl.h.
11175 (format_warning_va): Turned into small wrapper around
11176 format_warning_n_va, renamed to ...
11177 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
11178 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
11180 (format_warning_at_substring_n): New function.
11181 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
11182 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
11183 format_warning_at_substring with just a shorter name instead of
11184 const function pointer.
11185 (fmtwarn_n): New function.
11186 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
11187 appropriate, get rid of all the fmtstr temporaries, move conditionals
11188 with G_() wrapped string literals directly into fmtwarn arguments,
11189 cast dir.len to (int), formatting fixes.
11191 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
11193 * doc/invoke.texi: Remove "Cilk Plus" references.
11195 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11196 Richard Biener <rguenther@suse.de>
11199 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
11200 for error or warning attributes if CALL_FROM_THUNK_P is set.
11203 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11206 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
11207 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
11210 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
11211 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
11213 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
11214 instead of -1U in last predictors element's probability member.
11216 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
11219 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
11220 arguments if they are comparable.
11222 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11224 PR tree-optimization/84634
11225 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
11226 masks and masked_loop_p with a single loop_masks, making sure it's
11227 null for bb vectorization.
11229 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
11231 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11232 (vect_analyze_data_ref_access): Use loop->safe_len rather than
11233 loop->force_vectorize to check whether there is no alias.
11235 2018-03-02 Jakub Jelinek <jakub@redhat.com>
11238 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
11240 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
11242 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
11243 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
11244 instead of a loop around prev_real_insn.
11245 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
11248 PR inline-asm/84625
11249 * config/i386/i386.c (ix86_print_operand): Use conditional
11250 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
11253 2018-03-02 Richard Biener <rguenther@suse.de>
11255 PR tree-optimization/84427
11256 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
11257 (bitmap_set_subtract_values): Rewrite to handle multiple
11260 (prune_clobbered_mems): Likewise.
11261 (phi_translate): Take edge instead of pred/phiblock.
11262 (phi_translate_1): Likewise.
11263 (phi_translate_set): Likewise. Insert all translated
11264 exprs for a value into the set, keeping possibly multiple
11265 expressions per value.
11266 (compute_antic_aux): Adjust for phi_translate changes.
11267 When intersecting union the expressions and prune those
11268 not in the final value set, keeping possibly multiple
11269 expressions per value. Do not use value-insertion
11270 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
11271 all expressions. Add verification that the value-sets
11272 only shrink during iteration.
11273 (compute_partial_antic_aux): Adjust for the phi_translate changes.
11274 (do_pre_regular_insertion): Likewise.
11275 (do_pre_partial_partial_insertion): Likewise.
11277 2018-03-02 Richard Biener <rguenther@suse.de>
11280 * config/darwin.c (saved_debug_info_level): New static global.
11281 (darwin_asm_lto_start): Disable debug info generation for LTO out.
11282 (darwin_asm_lto_end): Restore debug info generation settings.
11284 2018-03-01 Martin Liska <mliska@suse.cz>
11287 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
11288 volatile arguments.
11290 2018-03-01 Richard Biener <rguenther@suse.de>
11293 * dwarf2out.c (gen_variable_die): Properly handle late VLA
11294 type annotation with LTO when debug was disabled at compile-time.
11296 2018-03-01 Matthew Fortune <mfortune@gmail.com>
11298 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
11300 (mips_final_postscan_insn): Likewise.
11302 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
11304 PR rtl-optimization/84528
11305 * alias.c (init_alias_target): Add commentary.
11306 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
11307 a unique base value if the frame pointer is not eliminated
11308 to the stack pointer.
11310 2018-03-01 Tom de Vries <tom@codesourcery.com>
11312 PR rtl-optimization/83327
11313 * lra-int.h (hard_regs_spilled_into): Declare.
11314 * lra.c (hard_regs_spilled_into): Define.
11315 (init_reg_info): Init hard_regs_spilled_into.
11316 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
11317 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
11318 (process_bb_lives): Handle hard_regs_spilled_into.
11319 (lra_create_live_ranges_1): Before doing liveness propagation, clear
11320 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
11322 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
11324 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
11325 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
11326 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
11327 * config/rs6000/aix72.h: New file.
11329 2018-02-28 Jakub Jelinek <jakub@redhat.com>
11331 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
11332 instead of warning_at with conditional singular and plural messages
11336 * stor-layout.c (update_alignment_for_field): For
11337 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
11338 && !DECL_PACKED (field), do the alignment update, just use
11339 only desired_align instead of MAX (type_align, desired_align)
11341 (place_field): Don't do known_align < desired_align handling
11342 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
11343 is non-NULL, instead do it after rli->prev_field handling and
11344 only if not within a bitfield word. For DECL_PACKED (field)
11345 use type_align of BITS_PER_UNIT.
11347 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
11349 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
11350 superfluous parentheses and trailing spaces.
11352 2018-02-28 Richard Biener <rguenther@suse.de>
11354 PR tree-optimization/84584
11355 * graphite-scop-detection.c (scop_detection::add_scop): Discard
11356 SCoPs with fake exit edge.
11358 2018-02-28 Martin Liska <mliska@suse.cz>
11361 * timevar.c (timer::print): Fix format to properly print 100%
11364 2018-02-28 Richard Biener <rguenther@suse.de>
11366 PR middle-end/84607
11367 * genmatch.c (capture_info::walk_match): Do not mark
11368 captured expressions without operands as expr_p given
11369 they act more like predicates and should be subject to
11370 "lost tail" side-effect preserving.
11372 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
11374 PR rtl-optimization/81611
11375 * auto-inc-dec.c (attempt_change): Move dead note from
11376 mem_insn if it's the next use of regno
11377 (find_address): Take address use of reg holding
11378 non-incremented value. Add parm to limit search to the named
11380 (merge_in_block): Attempt to use a mem insn that is the next
11381 use of the original regno.
11383 2018-02-27 Martin Sebor <msebor@redhat.com>
11386 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
11387 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
11389 2018-02-27 Martin Sebor <msebor@redhat.com>
11391 PR translation/84207
11392 * diagnostic-core.h (warning_n, error_n, inform_n): Change
11393 n argument to unsigned HOST_WIDE_INT.
11394 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
11395 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
11396 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
11397 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
11399 2018-02-27 Richard Biener <rguenther@suse.de>
11401 PR tree-optimization/84512
11402 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11403 Do not use the estimate returned from record_stmt_cost for
11404 the scalar iteration cost but sum properly using add_stmt_cost.
11406 2018-02-27 Richard Biener <rguenther@suse.de>
11408 PR tree-optimization/84466
11409 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11410 Adjust last change to less strictly validate use operands.
11412 2018-02-27 Martin Liska <mliska@suse.cz>
11414 PR gcov-profile/84548
11415 * gcov.c (process_file): Allow partial overlap and consider it
11416 also as group functions.
11417 (output_lines): Properly calculate range of lines for a group.
11419 2018-02-27 Martin Liska <mliska@suse.cz>
11421 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
11422 'ggc' suffixes. Change first column width.
11423 (timer::print): Fix formatting of the column.
11425 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
11427 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
11428 preserve inline entry blocks for the sake of debug inline
11429 entry point markers alone.
11430 (remove_unused_locals): Suggest in comments a better place to
11431 force the preservation of inline entry blocks that are
11432 otherwise unused, but do not preserve them.
11434 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11436 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
11438 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11441 * config/i386/constraints.md (Bs): Replace
11442 ix86_indirect_branch_register with
11443 TARGET_INDIRECT_BRANCH_REGISTER.
11445 * config/i386/i386.md (indirect_jump): Likewise.
11446 (tablejump): Likewise.
11447 (*sibcall_memory): Likewise.
11448 (*sibcall_value_memory): Likewise.
11449 Peepholes of indirect call and jump via memory: Likewise.
11450 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
11451 (*sibcall_value_GOT_32): Likewise.
11452 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11453 (GOT_memory_operand): Likewise.
11454 (call_insn_operand): Likewise.
11455 (sibcall_insn_operand): Likewise.
11456 (GOT32_symbol_operand): Likewise.
11457 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
11459 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11461 PR rtl-optimization/83496
11462 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
11463 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
11464 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
11465 redundant insn, if any.
11466 (relax_delay_slots): Likewise.
11467 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
11469 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
11471 PR tree-optimization/83965
11472 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
11473 that grouped statements are part of a reduction chain. Return
11474 true if the statement is not marked as a reduction itself but
11475 is part of a group.
11476 (vect_recog_dot_prod_pattern): Don't check whether the statement
11477 is part of a group here.
11478 (vect_recog_sad_pattern): Likewise.
11479 (vect_recog_widen_sum_pattern): Likewise.
11481 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
11484 * final.c (rest_of_clean_state): Also look for calls inside sequences.
11486 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
11489 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
11491 (ix86_output_indirect_function_return): New prototype.
11492 (ix86_split_simple_return_pop_internal): Likewise.
11493 * config/i386/i386.c (indirect_return_via_cx): New.
11494 (indirect_return_via_cx_bnd): Likewise.
11495 (indirect_thunk_name): Handle return va CX_REG.
11496 (output_indirect_thunk_function): Create alias for
11497 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
11498 (ix86_output_indirect_jmp): Remove the bool argument.
11499 (ix86_output_indirect_function_return): New function.
11500 (ix86_split_simple_return_pop_internal): Likewise.
11501 * config/i386/i386.md (*indirect_jump): Don't pass false
11502 to ix86_output_indirect_jmp.
11503 (*tablejump_1): Likewise.
11504 (simple_return_pop_internal): Change it to define_insn_and_split.
11505 Call ix86_split_simple_return_pop_internal to split it for
11506 -mfunction-return=.
11507 (simple_return_indirect_internal): Call
11508 ix86_output_indirect_function_return instead of
11509 ix86_output_indirect_jmp.
11511 2018-02-26 Jakub Jelinek <jakub@redhat.com>
11514 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
11515 memset and value initialization afterwards.
11517 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
11519 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
11521 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11524 * common/config/aarch64/aarch64-common.c
11525 (aarch_option_optimization_table[]): Switch
11526 off fomit-frame-pointer
11528 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11529 Chung-Ju Wu <jasonwucj@gmail.com>
11531 * config/nds32/nds32-multiple.md (load_multiple): Disallow
11533 (store_multiple): Ditto.
11535 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
11537 * config.gcc: Add --with-cpu support for nds32 target.
11538 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
11539 * config/nds32/nds32.opt: Add -mcpu= option.
11541 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
11543 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
11544 isel=yes): Warn for these deprecated options.
11546 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
11548 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
11549 ISA_2_5_MASKS_EMBEDDED.
11551 2018-02-23 Jakub Jelinek <jakub@redhat.com>
11553 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
11554 p->max as pointers rather than using iterative_hash_expr.
11556 2018-02-23 Carl Love <cel@us.ibm.com>
11558 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
11559 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11561 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
11562 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
11563 P8V_BUILTIN_VEC_VUNSIGNED2.
11565 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
11568 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
11569 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
11571 * lra-constraints.c (curr_insn_transform): Set up
11572 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
11574 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
11575 * lra-eliminations.c (spill_pseudos): Ditto.
11576 (process_insn_for_elimination): Ditto.
11577 * lra-lives.c (reg_early_clobber_p): Use the new macros.
11578 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
11579 LRA_NON_CLOBBERED_ALT.
11581 2018-02-22 Martin Sebor <msebor@redhat.com>
11583 PR tree-optimization/84480
11584 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
11585 to maybe_diag_stxncpy_trunc. Call it.
11586 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
11587 from gimple_fold_builtin_strcpy. Print inlining stack.
11588 (handle_builtin_stxncpy): Print inlining stack.
11589 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
11591 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
11594 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
11595 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
11596 and -fcheck-pointer-bounds are used together.
11597 (indirect_thunk_prefix): New enum.
11598 (indirect_thunk_need_prefix): New function.
11599 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
11600 "_nt" instead of "_bnd" for NOTRACK prefix.
11601 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
11602 (output_indirect_thunk_function): Likewise.
11604 (ix86_code_end): Update output_indirect_thunk_function calls.
11605 (ix86_output_indirect_branch_via_reg): Replace
11606 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
11607 (ix86_output_indirect_branch_via_push): Likewise.
11608 (ix86_output_function_return): Likewise.
11609 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
11610 incompatible with -fcf-protection=branch and
11611 -fcheck-pointer-bounds.
11613 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11616 * config/aarch64/aarch64.c (aarch64_print_address_internal):
11617 Change gcc_assert call to output_operand_lossage.
11619 2018-02-22 Steve Ellcey <sellcey@cavium.com>
11621 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
11623 2018-02-22 DJ Delorie <dj@redhat.com>
11624 Sebastian Perta <sebastian.perta@renesas.com>
11625 Oleg Endo <olegendo@gcc.gnu.org>
11627 * config/rx/rx.c (rx_rtx_costs): New function.
11628 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
11630 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
11632 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
11634 2018-02-22 Martin Liska <mliska@suse.cz>
11637 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
11638 Add "native" as a possible value.
11640 2018-02-22 Martin Liska <mliska@suse.cz>
11643 * config/i386/i386.c (ix86_option_override_internal):
11644 Add "native" as a possible value for -march and -mtune.
11646 2018-02-22 Jakub Jelinek <jakub@redhat.com>
11649 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
11650 to all type variants.
11652 PR tree-optimization/84503
11653 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
11654 width as info->bitpos + info->bitsize - start.
11655 (merged_store_group::merge_overlapping): Simplify width computation.
11656 (check_no_overlap): New function.
11657 (imm_store_chain_info::try_coalesce_bswap): Compute expected
11658 start + width and last_order of the group, fail if check_no_overlap
11660 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
11661 to group if check_no_overlap fails.
11663 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11665 * config/rs6000/altivec.md: Delete contraint arguments to
11666 define_expand, define_split, and define_peephole2, and in
11667 define_insn_and_split if always unused.
11668 * config/rs6000/darwin.md: Ditto.
11669 * config/rs6000/dfp.md: Ditto.
11670 * config/rs6000/rs6000.md: Ditto.
11671 * config/rs6000/sync.md: Ditto.
11672 * config/rs6000/vector.md: Ditto.
11673 * config/rs6000/vsx.md: Ditto.
11675 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
11677 * config/rs6000/altivec.md: Write output control strings as braced
11678 blocks instead of double-quoted strings.
11679 * config/rs6000/darwin.md: Ditto.
11680 * config/rs6000/rs6000.md: Ditto.
11681 * config/rs6000/vector.md: Ditto.
11682 * config/rs6000/vsx.md: Ditto.
11684 2018-02-21 Jason Merrill <jason@redhat.com>
11686 PR c++/84314 - ICE with templates and fastcall attribute.
11687 * attribs.c (build_type_attribute_qual_variant): Remove assert.
11689 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11691 * ipa-cp.c (determine_versionability): Fix comment typos.
11693 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
11696 * ipa-cp.c (determine_versionability): Do not version functions caling
11699 2018-02-21 Martin Liska <mliska@suse.cz>
11702 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
11703 Add "native" as a possible value.
11704 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
11705 the macro when native cpu detection is available.
11707 2018-02-21 Martin Liska <mliska@suse.cz>
11710 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
11711 Add "native" as a possible value.
11712 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
11713 when native cpu detection is available.
11715 2018-02-21 Jakub Jelinek <jakub@redhat.com>
11716 Martin Sebor <msebor@redhat.com>
11718 PR tree-optimization/84478
11719 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
11721 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
11722 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
11723 support which is conservatively correct, for 2 only stay conservative
11724 for maxlen. Formatting and comment capitalization fixes. Add STRICT
11725 argument to the 2 argument get_range_strlen, adjust 6 arg
11726 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
11728 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
11729 (gimple_fold_builtin_strlen): Pass true as last argument to
11732 2018-02-20 Martin Sebor <msebor@redhat.com>
11734 PR middle-end/84095
11735 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
11736 (builtin_memref::set_base_and_offset): Same. Handle inner references.
11737 (builtin_memref::builtin_memref): Factor out parts into
11738 set_base_and_offset and call it.
11740 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11742 PR middle-end/84406
11743 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
11744 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
11745 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
11746 search at the associated MODE_INT.
11748 2018-02-20 Jeff Law <law@redhat.com>
11750 PR middle-end/82123
11751 PR tree-optimization/81592
11752 PR middle-end/79257
11753 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
11754 for range data rather than using global data.
11755 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
11756 range data rather than using global data.
11757 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
11758 pass it to children as needed.
11759 (struct directive::fmtresult): Similarly.
11760 (struct directive::set_width): Similarly.
11761 (struct directive::set_precision): Similarly.
11762 (format_integer, format_directive, parse_directive): Similarly.
11763 (format_none): Accept unnamed vr_values parameter.
11764 (format_percent, format_floating, format_character): Similarly.
11765 (format_string, format_plain): Similarly.
11766 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
11767 the EVRP range analyzer for range data rather than using global data.
11768 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
11769 gimple-ssa-evrp-analyze.h
11770 (class sprintf_dom_walker): Add after_dom_children member function.
11771 Add evrp_range_analyzer member.
11772 (sprintf_dom_walker::before_dom_children): Call into the EVRP
11773 range analyzer as needed.
11774 (sprintf_dom_walker::after_dom_children): New member function.
11775 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
11777 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
11778 (evrp_range_analyzer::pop_to_marker): Likewise.
11780 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
11782 PR tree-optimization/84419
11783 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
11784 with the required type if its current type is compatible but
11787 2018-02-20 Jakub Jelinek <jakub@redhat.com>
11789 PR middle-end/82004
11790 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
11791 after vectorization.
11793 2018-02-20 Martin Liska <mliska@suse.cz>
11796 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
11797 possible values if we don't have a hint.
11799 2018-02-20 Martin Liska <mliska@suse.cz>
11803 * final.c (shorten_branches): Build align_tab array with one
11805 * opts.c (finish_options): Add alignment option limit check.
11806 (MAX_CODE_ALIGN): Likewise.
11807 (MAX_CODE_ALIGN_VALUE): Likewise.
11808 * doc/invoke.texi: Document maximum allowed option value for
11809 all -falign-* options.
11811 2018-02-19 Jakub Jelinek <jakub@redhat.com>
11814 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
11815 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
11816 * var-tracking.c (emit_note_insn_var_location): Remove all references
11817 to NOTE_INSN_CALL_ARG_LOCATION.
11818 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
11819 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
11820 Use copy_rtx_if_shared.
11821 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
11822 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
11823 (dwarf2out_var_location): Remove handling of
11824 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
11826 * final.c (final_scan_insn): Remove all references to
11827 NOTE_INSN_CALL_ARG_LOCATION.
11828 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
11829 before dumping final insns.
11830 * except.c (emit_note_eh_region_end): Remove all references to
11831 NOTE_INSN_CALL_ARG_LOCATION.
11832 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
11833 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
11834 * config/arc/arc.c (hwloop_optimize): Likewise.
11835 * config/arm/arm.c (create_fix_barrier): Likewise.
11836 * config/s390/s390.c (s390_chunkify_start): Likewise.
11837 * config/sh/sh.c (find_barrier): Likewise.
11838 * config/i386/i386.c (rest_of_insert_endbranch,
11839 ix86_seh_fixup_eh_fallthru): Likewise.
11840 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
11841 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
11842 * config/frv/frv.c (frv_function_prologue): Likewise.
11843 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
11845 (note_outside_basic_block_p): Remove all references to
11846 NOTE_INSN_CALL_ARG_LOCATION.
11847 * gengtype.c (adjust_field_rtx_def): Likewise.
11848 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
11850 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
11851 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
11854 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
11857 PR tree-optimization/84452
11858 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
11859 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
11862 2018-02-19 Martin Liska <mliska@suse.cz>
11865 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
11867 2018-02-19 Martin Liska <mliska@suse.cz>
11868 Richard Sandiford <richard.sandiford@linaro.org>
11870 PR tree-optimization/82491
11871 * gimple-fold.c (get_base_constructor): Make earlier bail out
11874 2018-02-19 Carl Love <cel@us.ibm.com>
11876 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
11877 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
11879 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
11880 P8V_BUILTIN_VEC_NEG.
11882 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
11884 * config/rl78/rl78.md (movdf): New define expand.
11886 2018-02-19 Martin Liska <mliska@suse.cz>
11889 * doc/invoke.texi: Fix typo.
11890 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
11892 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
11894 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
11895 handle rs6000_single_float and rs6000_double_float specially for
11898 2018-02-16 Jeff Law <law@redhat.com>
11900 * config/rx/rx.c (add_pop_cfi_notes): New function.;
11901 (pop_regs): Use it.
11903 2018-02-16 Jakub Jelinek <jakub@redhat.com>
11906 * ipa-inline.c (inline_small_functions): Fix a typo.
11908 2018-02-16 Nathan Sidwell <nathan@acm.org>
11910 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
11912 2018-02-16 Carl Love <cel@us.ibm.com>
11914 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
11915 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
11916 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
11917 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
11918 expansion to P8V_BUILTIN_VEC_FLOAT2.
11920 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
11922 PR rtl-optimization/70023
11923 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
11924 src_regno into account.
11926 2018-02-16 Carl Love <cel@us.ibm.com>
11928 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
11929 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
11930 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
11931 * config/rs6000/rs6000.c: Remove case statements for
11932 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
11933 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
11934 and P9V_BUILTIN_VEC_VINSERT4B.
11935 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
11936 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
11937 * config/rs6000/vsx.md:
11938 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
11941 2018-02-16 Carl Love <cel@us.ibm.com>
11943 * config/rs6000/altivec.h: Add builtin names vec_extract4b
11945 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
11947 * config/rs6000/rs6000-c.c: Add the definitions for
11948 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
11949 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
11950 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
11951 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
11952 definition for insert4b and define insn *insert3b_internal.
11953 * doc/extend.texi: Add documentation for vec_extract4b.
11955 2018-02-16 Nathan Sidwell <nathan@acm.org>
11957 * doc/extend.texi (Backwards Compatibility): Mention friend
11958 injection. Note for-scope is deprecated.
11959 * doc/invoke.texi (-ffriend-injection): Deprecate.
11961 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
11963 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
11964 that moved to I2, also allow destinations that are a paradoxical
11965 subreg (instead of a normal reg).
11967 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
11970 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
11973 2018-02-16 Richard Biener <rguenther@suse.de>
11975 PR tree-optimization/84037
11976 PR tree-optimization/84016
11978 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11979 Adjust vec_construct for the fact we need additional higher latency
11980 128bit inserts for AVX256 and AVX512 vector builds.
11981 (ix86_add_stmt_cost): Scale vector construction cost for
11984 2018-02-16 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/84417
11987 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
11988 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
11989 (non_rewritable_lvalue_p): Likewise, use poly-ints.
11991 2018-02-16 Martin Liska <mliska@suse.cz>
11994 * internal-fn.def (ASAN_CHECK): Set proper flags.
11995 (ASAN_MARK): Likewise.
11997 2018-02-16 Julia Koval <julia.koval@intel.com>
11999 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12000 from PTA_CANNONLAKE.
12002 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12005 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12006 Use ++iter rather than iter++ for std::list iterators.
12007 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12008 defer deleting them until all nodes in the forest are processed. Do
12009 free even leaf nodes. Change to_process into auto_vec.
12012 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12013 * vec.h (vec_default_construct): Use memset instead of placement new
12014 if BROKEN_VALUE_INITIALIZATION is defined.
12015 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12016 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12019 PR rtl-optimization/83723
12020 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12021 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12022 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12024 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12026 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12028 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12030 PR rtl-optimization/81443
12031 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12032 from inner REGs to paradoxical SUBREGs.
12034 2018-02-16 Richard Biener <rguenther@suse.de>
12036 PR tree-optimization/84399
12037 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12038 For operands we can analyze at their definition make sure we can
12039 analyze them at each use as well.
12041 2018-02-16 Richard Biener <rguenther@suse.de>
12043 PR tree-optimization/84190
12044 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12045 volatile accesses if the decl isn't volatile.
12047 2018-02-15 Jason Merrill <jason@redhat.com>
12049 PR c++/84314 - ICE with templates and fastcall attribute.
12050 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12051 TYPE_CANONICAL on an existing type.
12053 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12055 PR tree-optimization/84383
12056 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12057 dstoff nor call operand_equal_p if dstbase is NULL.
12059 PR tree-optimization/84334
12060 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12061 also a CONSTANT_CLASS_P, punt.
12063 2018-02-14 Jim Wilson <jimw@sifive.com>
12065 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
12066 first SMALL_OPERAND check. New local min_second_step. Move assert
12067 to where locals are set. Add TARGET_RVC support.
12068 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
12070 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
12072 * doc/invoke.texi: Correct -Wformat-overflow code sample.
12074 2018-02-14 Martin Sebor <msebor@redhat.com>
12076 PR tree-optimization/83698
12077 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
12078 arrays constrain the offset range to their bounds.
12079 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
12080 (builtin_access::overlap): Avoid setting the size of overlap if it's
12082 (maybe_diag_overlap): Also consider arrays when deciding what values
12083 of offsets to include in diagnostics.
12085 2018-02-14 Martin Sebor <msebor@redhat.com>
12088 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
12089 that correspond to the kind of a declaration.
12091 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
12094 * config/pa/pa.md: Load address of PIC label using the linkage table
12095 if the label is nonlocal.
12097 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
12099 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
12100 warning message if user requests -maltivec=be.
12101 * doc/invoke.texi: Document deprecation of -maltivec=be.
12103 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
12106 * config/rs6000/rs6000-c.c: Update definitions for
12107 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
12108 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
12110 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
12113 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
12114 add _get_ssp intrinsics. Remove argument from
12115 __builtin_ia32_rdssp[d|q].
12116 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
12117 * config/i386/i386-builtin.def: Remove argument from
12118 __builtin_ia32_rdssp[d|q].
12119 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
12120 ix86_expand_special_args_builtin for _rdssp[d|q].
12121 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
12122 Clear register before usage.
12123 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
12124 Add documentation for new _get_ssp and _inc_ssp intrinsics.
12126 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
12128 PR tree-optimization/84357
12129 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
12130 operand 1 of an ARRAY_REF too.
12132 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
12135 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
12136 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12138 (set_of_reg): New struct.
12139 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
12140 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
12141 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
12143 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
12144 Split into bitclr, bitset, bitinvert patterns if appropriate.
12145 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
12146 use rx_fuse_in_memory_bitop.
12147 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
12148 to named insn, correct maximum insn length.
12150 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
12153 * machmode.def: Define a complex mode for PARTIAL_INT.
12154 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
12156 * doc/rtl.texi: Document CSPImode.
12157 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
12159 (msp430_hard_regno_nregs_with_padding): Likewise.
12161 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
12164 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
12166 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
12168 PR rtl-optimization/84169
12169 * combine.c (try_combine): New variable split_i2i3. Set it to true if
12170 we generated a parallel as new i3 and we split that to new i2 and i3
12171 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
12172 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
12173 those to i2, not i1. Partially rewrite this scan code.
12175 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12178 * stor-layout.c (place_field): For variable length fields, adjust
12179 offset_align afterwards not just based on the field's alignment,
12180 but also on the size.
12182 PR middle-end/84309
12183 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
12184 of exps and logs in the use_exp2 case.
12186 2018-02-13 Jeff Law <law@redhat.com>
12188 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
12189 entry for "vector".
12191 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
12194 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
12198 * common.opt (gas-loc-support, gas-locview-support): New.
12199 (ginline-points, ginternal-reset-location-views): New.
12200 * doc/invoke.texi: Document them. Use @itemx where intended.
12201 (gvariable-location-views): Adjust.
12202 * target.def (reset_location_view): New.
12203 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
12204 (TARGET_RESET_LOCATION_VIEW): New.
12205 * doc/tm.texi: Rebuilt.
12206 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
12207 (dwarf2out_default_as_locview_support): New.
12208 (output_asm_line_debug_info): Use option variables.
12209 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
12210 (output_loc_list): Likewise.
12211 (add_high_low_attributes): Check option variables.
12212 Don't output entry view attribute in strict mode.
12213 (gen_inlined_subroutine_die): Check option variables.
12214 (dwarf2out_inline_entry): Likewise.
12215 (init_sections_and_labels): Likewise.
12216 (dwarf2out_early_finish): Likewise.
12217 (maybe_reset_location_view): New, from...
12218 (dwarf2out_var_location): ... here. Call it.
12219 * debug.h (dwarf2out_default_as_loc_support): Declare.
12220 (dwarf2out_default_as_locview_support): Declare.
12221 * hooks.c (hook_int_rtx_insn_0): New.
12222 * hooks.h (hook_int_rtx_insn_0): Declare.
12223 * toplev.c (process_options): Take -gas-loc-support and
12224 -gas-locview-support from dwarf2out. Enable
12225 -gvariable-location-views by default only with locview
12226 assembler support. Enable -ginternal-reset-location-views by
12227 default only if the target defines the corresponding hook.
12228 Enable -ginline-points by default if location views are
12229 enabled; force it disabled if statement frontiers are
12231 * tree-inline.c (expand_call_inline): Check option variables.
12232 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
12234 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
12236 PR tree-optimization/84321
12237 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
12238 handling. Also check whether the anti-range contains any values
12239 that satisfy the mask; switch to a VR_RANGE if not.
12241 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
12244 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
12246 2018-02-13 Martin Jambor <mjambor@suse.cz>
12249 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
12250 of call statements, also set location of a load to a temporary.
12252 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
12254 * config/rl78/rl78.c (add_vector_labels): New function.
12255 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
12256 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
12257 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
12258 which checks that no arguments are passed.
12259 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
12260 * doc/extend.texi: Documentation for the new attribute.
12262 2018-02-13 Andreas Schwab <schwab@suse.de>
12264 * config/riscv/linux.h (CPP_SPEC): Define.
12266 2018-02-13 Jakub Jelinek <jakub@redhat.com>
12269 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
12270 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
12271 OPTION_MASK_ISA_AES as first argument to def_builtin_const
12272 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
12273 instead of OPTION_MASK_ISA_PCLMUL as first argument to
12274 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
12275 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
12276 temporarily for AES and PCLMUL builtins.
12278 PR tree-optimization/84339
12279 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
12280 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
12283 PR middle-end/84309
12284 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
12285 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
12286 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
12288 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
12290 * omp-simd-clone.h: New file.
12291 * omp-simd-clone.c: Include omp-simd-clone.h.
12292 (expand_simd_clones): No longer static.
12293 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
12294 cgraph.h and omp-simd-clone.h.
12295 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
12296 (vect_recog_widen_shift_pattern): Formatting fix.
12297 (vect_pattern_recog_1): Don't check optab for calls.
12300 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
12301 operands[2] into a REG before using gen_lowpart on it.
12303 2018-02-12 Jeff Law <law@redhat.com>
12306 * config/sh/sh.c (find_barrier): Consider a sibling call
12309 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
12310 successfully back substituting a reg.
12312 2018-02-12 Richard Biener <rguenther@suse.de>
12314 PR tree-optimization/84037
12315 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
12316 parameter, move visited init to caller.
12317 (vect_slp_analyze_operations): Separate cost from validity
12318 check, initialize visited once for all instances.
12319 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
12321 * tree-vect-stmts.c (vect_model_simple_cost): Make early
12323 (vect_model_promotion_demotion_cost): Likewise.
12324 (vectorizable_bswap): Guard cost modeling with !slp_node
12325 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
12327 (vectorizable_call): Likewise.
12328 (vectorizable_conversion): Likewise.
12329 (vectorizable_assignment): Likewise.
12330 (vectorizable_shift): Likewise.
12331 (vectorizable_operation): Likewise.
12332 (vectorizable_store): Likewise.
12333 (vectorizable_load): Likewise.
12334 (vectorizable_condition): Likewise.
12335 (vectorizable_comparison): Likewise.
12337 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
12340 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
12341 (ASAN_MARK): Fix fnspec to account for return value, change pointer
12342 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
12344 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12346 PR middle-end/83665
12347 * params.def (inline-min-speedup): Increase from 8 to 15.
12348 (max-inline-insns-auto): Decrease from 40 to 30.
12349 * ipa-split.c (consider_split): Add some buffer for function to
12350 be considered inlining candidate.
12351 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
12354 2018-02-12 Richard Biener <rguenther@suse.de>
12356 PR tree-optimization/84037
12357 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
12358 matched stmts if we cannot swap the non-matched ones.
12360 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12362 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
12363 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
12364 _mm_maskz_scalef_round_ss): New intrinsics.
12365 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
12366 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
12367 __builtin_ia32_scalefss_round): Remove.
12368 (__builtin_ia32_scalefsd_mask_round,
12369 __builtin_ia32_scalefss_mask_round): New intrinsics.
12370 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
12371 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
12372 ((match_operand:VF_128 2 "<round_nimm_predicate>"
12373 "<round_constraint>")): Changed to ...
12374 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
12375 "<round_scalar_constraint>")): ... this.
12376 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
12377 %0, %1, %2<round_op3>}"): Changed to ...
12378 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
12379 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
12380 %2<round_scalar_mask_op3>}"): ... this.
12381 * config/i386/subst.md (round_scalar_nimm_predicate): New.
12383 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
12385 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
12386 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
12387 (_mm_maskz_sqrt_round_ss): New intrinsics.
12388 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
12389 (__builtin_ia32_sqrtsd_mask_round)
12390 (__builtin_ia32_sqrtss_mask_round): New builtins.
12391 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
12392 (__builtin_ia32_sqrtss_round): Remove.
12393 (__builtin_ia32_sqrtsd_mask_round)
12394 (__builtin_ia32_sqrtss_mask_round): New builtins.
12395 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
12396 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
12397 ((match_operand:VF_128 1 "vector_operand"
12398 "xBm,<round_constraint>")): Changed to ...
12399 ((match_operand:VF_128 1 "vector_operand"
12400 "xBm,<round_scalar_constraint>")): ... this.
12401 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
12402 %0, %2, %<iptr>1<round_op3>}): Changed to ...
12403 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
12404 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
12405 %<iptr>1<round_scalar_mask_op3>}): ... this.
12406 ((set_attr "prefix" "<round_prefix>")): Changed to ...
12407 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
12409 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
12412 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
12413 Cast vec_cmpeq result to correct type.
12414 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
12415 Cast vec_cmpgt result to correct type.
12417 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
12419 * final.c (final_scan_insn_1): Renamed from...
12420 (final_scan_insn): ... this. New wrapper, to recover
12421 seen from the outermost call in recursive ones.
12422 * config/sparc/sparc.c (output_return): Drop seen from call.
12423 (output_sibcall): Likewise.
12424 * config/visium/visium.c (output_branch): Likewise.
12426 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
12428 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
12431 2018-02-10 Alan Modra <amodra@gmail.com>
12434 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
12435 Specify LR as an input.
12437 2018-02-10 Jakub Jelinek <jakub@redhat.com>
12440 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
12441 remove_member_access_dummy_vars): New functions.
12442 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
12443 lower_omp_1, execute_lower_omp): Use them.
12445 PR rtl-optimization/84308
12446 * shrink-wrap.c (spread_components): Release todo vector.
12448 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
12450 PR rtl-optimization/57193
12451 * ira-color.c (struct allocno_color_data): Add member
12452 conflict_allocno_hard_prefs.
12453 (update_conflict_allocno_hard_prefs): New.
12454 (bucket_allocno_compare_func): Add a preference based on
12455 conflict_allocno_hard_prefs.
12456 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
12457 (color_allocnos): Remove a dead code. Initiate
12458 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
12460 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12463 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
12464 constraint from =wa to wa. Avoid a subreg on the output operand,
12465 instead use a pseudo and subreg it in a move.
12466 (p9_xxbrd_<mode>): Changed to ...
12467 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
12468 (p9_xxbrd_v2df): New expander.
12469 (p9_xxbrw_<mode>): Changed to ...
12470 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
12471 (p9_xxbrw_v4sf): New expander.
12473 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12475 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
12477 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
12480 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
12481 multiply in 32-bit mode.
12482 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
12483 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
12486 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
12488 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
12489 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
12490 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
12491 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
12493 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
12496 * dwarf2out.c (is_trivial_indirect_ref): New function.
12497 (dwarf2out_late_global_decl): Do not generate a location
12498 attribute for variables that have a non-trivial DECL_VALUE_EXPR
12499 and that are not defined in the current unit.
12501 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12503 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
12504 instead of a libcall for UNORDERED.
12506 2018-02-09 Tamar Christina <tamar.christina@arm.com>
12509 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
12510 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
12512 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12515 * config/s390/s390.c (s390_set_current_function): Invoke
12516 s390_indirect_branch_settings also if fndecl didn't change.
12518 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12520 * config/rs6000/rs6000.md (blockage): Set length to zero.
12522 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
12524 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
12526 2018-02-09 Jakub Jelinek <jakub@redhat.com>
12529 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
12530 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
12534 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
12535 PARALLEL incoming that failed vt_get_decl_and_offset check.
12537 PR middle-end/84237
12538 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
12539 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
12541 (get_variable_section): For decls in named .bss* sections pass true as
12542 second argument to bss_initializer_p.
12544 2018-02-09 Marek Polacek <polacek@redhat.com>
12545 Jakub Jelinek <jakub@redhat.com>
12548 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
12549 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
12550 Sync some changes from cxx_fold_indirect_ref.
12552 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
12554 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
12556 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
12557 (BLOCK_INLINE_ENTRY_LABEL): New.
12558 (dwarf2out_var_location): Disregard inline entry markers.
12559 (inline_entry_data): New struct.
12560 (inline_entry_data_hasher): New hashtable type.
12561 (inline_entry_data_hasher::hash): New.
12562 (inline_entry_data_hasher::equal): New.
12563 (inline_entry_data_table): New variable.
12564 (add_high_low_attributes): Add DW_AT_entry_pc and
12565 DW_AT_GNU_entry_view attributes if a pending entry is found
12566 in inline_entry_data_table. Add old entry_pc attribute only
12567 if debug nonbinding markers are disabled.
12568 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
12569 markers are enabled.
12570 (block_within_block_p, dwarf2out_inline_entry): New.
12571 (dwarf2out_finish): Check that no entries remained in
12572 inline_entry_data_table.
12573 * final.c (reemit_insn_block_notes): Handle inline entry notes.
12574 (final_scan_insn, notice_source_line): Likewise.
12575 (rest_of_clean_state): Skip inline entry markers.
12576 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
12578 * gimple.c (gimple_build_debug_inline_entry): New.
12579 * gimple.h (enum gimple_debug_subcode): Add
12580 GIMPLE_DEBUG_INLINE_ENTRY.
12581 (gimple_build_debug_inline_entry): Declare.
12582 (gimple_debug_inline_entry_p): New.
12583 (gimple_debug_nonbind_marker_p): Adjust.
12584 * insn-notes.def (INLINE_ENTRY): New.
12585 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
12586 inline entry marker notes.
12587 (print_insn): Likewise.
12588 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
12589 (INSN_DEBUG_MARKER_KIND): Likewise.
12590 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
12591 * tree-inline.c (expand_call_inline): Build and insert
12592 debug_inline_entry stmt.
12593 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
12594 inline entry blocks early, if nonbind markers are enabled.
12595 (dump_scope_block): Dump fragment info.
12596 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
12597 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
12598 (gimple_build_debug_inline_entry): New.
12599 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
12600 Enable/disable inline entry points too.
12601 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
12602 (DEBUG_INSN): Describe inline entry markers.
12604 * common.opt (gvariable-location-views): New.
12605 (gvariable-location-views=incompat5): New.
12606 * config.in: Rebuilt.
12607 * configure: Rebuilt.
12608 * configure.ac: Test assembler for view support.
12609 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
12610 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
12611 * dwarf2out.c (var_loc_view): New typedef.
12612 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
12613 (dwarf2out_locviews_in_attribute): New.
12614 (dwarf2out_locviews_in_loclist): New.
12615 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
12616 (enum dw_line_info_opcode): Add LI_adv_address.
12617 (struct dw_line_info_table): Add view.
12618 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
12619 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
12620 (zero_view_p): New variable.
12621 (ZERO_VIEW_P): New macro.
12622 (output_asm_line_debug_info): New.
12623 (struct var_loc_node): Add view.
12624 (add_AT_view_list, AT_loc_list): New.
12625 (add_var_loc_to_decl): Add view param. Test it against last.
12626 (new_loc_list): Add view params. Record them.
12627 (AT_loc_list_ptr): Handle loc and view lists.
12628 (view_list_to_loc_list_val_node): New.
12629 (print_dw_val): Handle dw_val_class_view_list.
12630 (size_of_die): Likewise.
12631 (value_format): Likewise.
12632 (loc_list_has_views): New.
12633 (gen_llsym): Set vl_symbol too.
12634 (maybe_gen_llsym, skip_loc_list_entry): New.
12635 (dwarf2out_maybe_output_loclist_view_pair): New.
12636 (output_loc_list): Output view list or entries too.
12637 (output_view_list_offset): New.
12638 (output_die): Handle dw_val_class_view_list.
12639 (output_dwarf_version): New.
12640 (output_compilation_unit_header): Use it.
12641 (output_skeleton_debug_sections): Likewise.
12642 (output_rnglists, output_line_info): Likewise.
12643 (output_pubnames, output_aranges): Update version comments.
12644 (output_one_line_info_table): Output view numbers in asm comments.
12645 (dw_loc_list): Determine current endview, pass it to new_loc_list.
12646 Call maybe_gen_llsym.
12647 (loc_list_from_tree_1): Adjust.
12648 (add_AT_location_description): Create view list attribute if
12649 needed, check it's absent otherwise.
12650 (convert_cfa_to_fb_loc_list): Adjust.
12651 (maybe_emit_file): Call output_asm_line_debug_info for test.
12652 (dwarf2out_var_location): Reset views as needed. Precompute
12653 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
12654 attribute. Set view.
12655 (new_line_info_table): Reset next view.
12656 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
12657 (dwarf2out_source_line): Likewise. Output view resets and labels to
12658 the assembler, or select appropriate line info opcodes.
12659 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
12660 (optimize_string_length): Catch it. Adjust.
12661 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
12662 dw_val_class_view_list, and remove it if no longer needed.
12663 (hash_loc_list): Hash view numbers.
12664 (loc_list_hasher::equal): Compare them.
12665 (optimize_location_lists): Check whether a view list symbol is
12666 needed, and whether the locview attribute is present, and
12667 whether they match. Remove the locview attribute if no longer
12669 (index_location_lists): Call skip_loc_list_entry for test.
12670 (dwarf2out_finish): Call output_asm_line_debug_info for test.
12671 Use output_dwarf_version.
12672 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
12673 (struct dw_val_node): Add val_view_list.
12674 * final.c (SEEN_NEXT_VIEW): New.
12675 (set_next_view_needed): New.
12676 (clear_next_view_needed): New.
12677 (maybe_output_next_view): New.
12678 (final_start_function): Rename to...
12679 (final_start_function_1): ... this. Take pointer to FIRST,
12680 add SEEN parameter. Emit param bindings in the initial view.
12681 (final_start_function): Reintroduce SEEN-less interface.
12682 (final): Rename to...
12683 (final_1): ... this. Take SEEN parameter. Output final pending
12684 next view at the end.
12685 (final): Reintroduce seen-less interface.
12686 (final_scan_insn): Output pending next view before switching
12687 sections or ending a block. Mark the next view as needed when
12688 outputting variable locations. Notify debug backend of section
12689 changes, and of location view changes.
12690 (rest_of_handle_final): Adjust.
12691 * toplev.c (process_options): Autodetect value for debug variable
12692 location views option. Warn on incompat5 without -gdwarf-5.
12693 * doc/invoke.texi (gvariable-location-views): New.
12694 (gvariable-location-views=incompat5): New.
12695 (gno-variable-location-views): New.
12697 2018-02-08 David Malcolm <dmalcolm@redhat.com>
12699 PR tree-optimization/84136
12700 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
12701 that the result of find_edge is non-NULL.
12703 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12706 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
12707 storing integer register in SImode. Fix cost of 256 and 512
12708 byte aligned SSE register store.
12710 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
12712 * config/i386/i386.c (ix86_multiplication_cost): Fix
12713 multiplication cost for TARGET_AVX512DQ.
12715 2018-02-08 Marek Polacek <polacek@redhat.com>
12717 PR tree-optimization/84238
12718 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
12721 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12723 PR tree-optimization/84265
12724 * tree-vect-stmts.c (vectorizable_store): Don't treat
12725 VMAT_CONTIGUOUS accesses as grouped.
12726 (vectorizable_load): Likewise.
12728 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12730 PR tree-optimization/81635
12731 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
12732 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
12733 (test_round_for_mask): New functions.
12734 (wide_int_cc_tests): Call test_round_for_mask.
12735 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
12736 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
12737 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
12738 range returned by get_range_info.
12740 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
12743 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
12744 * symtab.c: Include builtins.h
12745 (symtab_node::output_to_lto_symbol_table_p): Move here
12746 from lto-streamer-out.c:output_symbol_p.
12747 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
12748 (output_symbol_p): Move all logic to symtab.c
12749 (produce_symtab): Update.
12751 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12753 * config/s390/s390-opts.h (enum indirect_branch): Define.
12754 * config/s390/s390-protos.h (s390_return_addr_from_memory)
12755 (s390_indirect_branch_via_thunk)
12756 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
12757 (enum s390_indirect_branch_type): Define.
12758 * config/s390/s390.c (struct s390_frame_layout, struct
12759 machine_function): Remove.
12760 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
12761 (indirect_branch_table_label_no, indirect_branch_table_name):
12763 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
12764 (enum s390_indirect_branch_option): Define.
12765 (s390_return_addr_from_memory): New function.
12766 (s390_handle_string_attribute): New function.
12767 (s390_attribute_table): Add new attribute handler.
12768 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
12769 (s390_indirect_branch_via_thunk): New function.
12770 (s390_indirect_branch_via_inline_thunk): New function.
12771 (s390_function_ok_for_sibcall): When jumping via thunk disallow
12772 sibling call optimization for non z10 compiles.
12773 (s390_emit_call): Force indirect branch target to be a single
12774 register. Add r1 clobber for non-z10 compiles.
12775 (s390_emit_epilogue): Emit return jump via return_use expander.
12776 (s390_reorg): Handle JUMP_INSNs as execute targets.
12777 (s390_option_override_internal): Perform validity checks for the
12778 new command line options.
12779 (s390_indirect_branch_attrvalue): New function.
12780 (s390_indirect_branch_settings): New function.
12781 (s390_set_current_function): Invoke s390_indirect_branch_settings.
12782 (s390_output_indirect_thunk_function): New function.
12783 (s390_code_end): Implement target hook.
12784 (s390_case_values_threshold): Implement target hook.
12785 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
12787 * config/s390/s390.h (struct s390_frame_layout)
12788 (struct machine_function): Move here from s390.c.
12789 (TARGET_INDIRECT_BRANCH_NOBP_RET)
12790 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
12791 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
12792 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
12793 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
12794 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
12795 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
12796 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
12797 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
12798 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
12799 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
12800 (mnemonic attribute): Add values which aren't recognized
12802 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
12803 pattern for branch conversion. Fix mnemonic attribute.
12804 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
12805 indirect branch via thunk if requested.
12806 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
12807 ("*indirect_jump"): Disable for branch conversion using out of
12809 ("indirect_jump_via_thunk<mode>_z10")
12810 ("indirect_jump_via_thunk<mode>")
12811 ("indirect_jump_via_inlinethunk<mode>_z10")
12812 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
12813 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
12814 ("casesi_jump_via_inlinethunk<mode>_z10")
12815 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
12816 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
12817 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
12818 ("*indirect2_jump"): Disable for branch conversion.
12819 ("casesi_jump"): Turn into expander and expand patterns for branch
12821 ("return_use"): New expander.
12822 ("*return"): Emit return via thunk and rename it to ...
12823 ("*return<mode>"): ... this one.
12824 * config/s390/s390.opt: Add new options and and enum for the
12827 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
12829 * lra-constraints.c (match_reload): Unconditionally use
12830 gen_lowpart_SUBREG, rather than selecting between that
12831 and equivalent gen_rtx_SUBREG code.
12833 2018-02-08 Richard Biener <rguenther@suse.de>
12835 PR tree-optimization/84233
12836 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
12837 changed flag instead of boguously re-using phi_inserted.
12839 2018-02-08 Martin Jambor <mjambor@suse.cz>
12841 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
12842 static local variables.
12844 2018-02-08 Richard Biener <rguenther@suse.de>
12846 PR tree-optimization/84278
12847 * tree-vect-stmts.c (vectorizable_store): When looking for
12848 smaller vector types to perform grouped strided loads/stores
12849 make sure the mode is supported by the target.
12850 (vectorizable_load): Likewise.
12852 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12854 * config/aarch64/aarch64.c (aarch64_components_for_bb):
12855 Increase LDP/STP opportunities by adding adjacent callee-saves.
12857 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
12859 PR rtl-optimization/84068
12860 PR rtl-optimization/83459
12861 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
12863 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
12865 PR tree-optimization/84224
12866 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
12867 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
12868 non-zero arguments.
12870 2018-02-07 Iain Sandoe <iain@codesourcery.com>
12873 * config/rs6000/altivec.md (*restore_world): Remove LR use.
12874 * config/rs6000/predicates.md (restore_world_operation): Adjust op
12875 count, remove one USE.
12877 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
12879 * doc/install.texi (Configuration): Document the
12880 --with-long-double-format={ibm,ieee} PowerPC configuration
12884 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
12885 Convert from define_expand to be define_insn_and_split. Rework
12886 float/double/_Float128 conversions to QI/HI/SImode to work with
12887 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
12888 conversions to QI/HImode types did a store and then a load to
12889 truncate the value. For conversions to VSX registers, don't split
12890 the insn, instead emit the code directly. Use the code iterator
12891 any_fix to combine signed and unsigned conversions.
12892 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
12893 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
12894 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12895 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
12896 (fix_<mode>di2_hw): Likewise.
12897 (fixuns_<mode>di2_hw): Likewise.
12898 (fix_<mode>si2_hw): Likewise.
12899 (fixuns_<mode>si2_hw): Likewise.
12900 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
12901 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
12902 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
12903 fix<uns>_trunc<SFDF:mode>si2_p8.
12904 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
12905 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
12906 (fix<uns>_<mode>_mem): Likewise.
12907 (fctiw<u>z_<mode>_mem): Likewise.
12908 (fix<uns>_<mode>_mem): Likewise.
12909 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
12910 the register allocator from doing a direct move to the GPRs to do
12911 a store, and instead use the ISA 3.0 store byte/half-word from
12912 vector register instruction. For IEEE 128-bit floating point,
12913 also optimize stores of 32-bit ints.
12914 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
12916 2018-02-07 Alan Hayward <alan.hayward@arm.com>
12918 * genextract.c (push_pathstr_operand): New function to support
12920 (walk_rtx): Call push_pathstr_operand.
12921 (print_path): Support [a-zA-Z].
12923 2018-02-07 Richard Biener <rguenther@suse.de>
12925 PR tree-optimization/84037
12926 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
12927 (cse_and_gimplify_to_preheader): Declare.
12928 (vect_get_place_in_interleaving_chain): Likewise.
12929 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12931 (_loop_vec_info::~_loop_vec_info): Delete it.
12932 (cse_and_gimplify_to_preheader): New function.
12933 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
12934 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
12935 (vectorizable_load): Likewise. For grouped stores always base
12936 the IV on the first element.
12937 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
12938 condition before gimplifying.
12940 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12942 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
12943 *DIV_EXPR and *MOD_EXPR.
12945 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
12948 * config/i386/i386.c (ix86_option_override_internal): Mask out
12949 the CF_SET bit when checking -fcf-protection.
12951 2018-02-07 Tom de Vries <tom@codesourcery.com>
12954 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
12957 2018-02-07 Richard Biener <rguenther@suse.de>
12959 PR tree-optimization/84204
12960 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
12963 PR tree-optimization/84205
12964 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
12965 special-case isl_ast_op_zdiv_r.
12967 PR tree-optimization/84223
12968 * graphite-scop-detection.c (gather_bbs::before_dom_children):
12969 Only add conditions from within the region.
12970 (gather_bbs::after_dom_children): Adjust.
12972 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
12975 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
12976 * config/avr/avr.md: Only post-reload split REG-REG moves if
12977 either register is GENERAL_REG_P.
12979 2018-02-07 Jakub Jelinek <jakub@redhat.com>
12981 PR tree-optimization/84235
12982 * tree-ssa-scopedtables.c
12983 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
12984 if the subtraction is performed in floating point type where NaNs are
12985 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
12986 build 1. Formatting fix.
12988 2018-02-06 Jakub Jelinek <jakub@redhat.com>
12991 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
12992 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
12993 and skip it regardless of bb boundaries. Use CALL_P macro,
12994 don't test INSN_P (insn) together with CALL_P or JUMP_P check
12995 unnecessarily, formatting fix.
12997 2018-02-06 Michael Collison <michael.collison@arm.com>
12999 * config/arm/thumb2.md:
13000 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13001 (*thumb_mov_notscc): Ditto.
13003 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13006 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13007 unsigned_fix, not "s".
13009 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13011 * configure.ac (gcc_fn_eh_frame_ro): New function.
13012 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13013 correct .eh_frame permissions.
13014 * configure: Regenerate.
13016 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13018 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13019 irrelevant options.
13021 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13023 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13024 Display warning message for -mno-speculate-indirect-jumps.
13026 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13028 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13030 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13032 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13034 PR tree-optimization/84225
13035 * tree-eh.c (find_trapping_overflow): Only call
13036 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13038 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13041 * config/i386/i386.c: Reimplement the check of possible options
13042 -mibt/-mshstk conbination. Change error messages.
13043 * doc/invoke.texi: Fix a typo: remove extra '='.
13045 2018-02-06 Marek Polacek <polacek@redhat.com>
13047 PR tree-optimization/84228
13048 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13050 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13053 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13054 emitted arch directives.
13055 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13056 __ARM_FEATURE_COPROC before changing architectures.
13058 2018-02-06 Richard Biener <rguenther@suse.de>
13060 * config/i386/i386.c (print_reg): Fix typo.
13061 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
13063 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
13065 * configure: Regenerate.
13067 2018-02-05 Martin Sebor <msebor@redhat.com>
13069 PR tree-optimization/83369
13070 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
13073 2018-02-05 Martin Liska <mliska@suse.cz>
13075 * doc/invoke.texi: Cherry-pick upstream r323995.
13077 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
13079 * ira.c (ira_init_register_move_cost): Adjust comment.
13081 2018-02-05 Martin Liska <mliska@suse.cz>
13083 PR gcov-profile/84137
13084 * doc/gcov.texi: Fix typo in documentation.
13086 2018-02-05 Martin Liska <mliska@suse.cz>
13088 PR gcov-profile/83879
13089 * doc/gcov.texi: Document necessity of --dynamic-list-data when
13090 using dlopen functionality.
13092 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
13094 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
13095 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
13096 _mm_maskz_range_ss, _mm_mask_range_round_ss,
13097 _mm_maskz_range_round_ss): New intrinsics.
13098 (__builtin_ia32_rangesd128_round)
13099 (__builtin_ia32_rangess128_round): Remove.
13100 (__builtin_ia32_rangesd128_mask_round,
13101 __builtin_ia32_rangess128_mask_round): New builtins.
13102 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
13103 __builtin_ia32_rangess128_round): Remove.
13104 (__builtin_ia32_rangesd128_mask_round,
13105 __builtin_ia32_rangess128_mask_round): New builtins.
13106 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
13107 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
13108 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
13109 "<round_saeonly_constraint>")): Changed to ...
13110 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
13111 "<round_saeonly_scalar_constraint>")): ... this.
13112 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
13113 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
13114 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
13115 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
13116 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
13118 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
13120 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
13122 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
13123 Remove all values except native, 8540 and 8548.
13125 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
13127 * config/i386/i386.c (ix86_output_function_return): Pass
13128 INVALID_REGNUM, instead of -1, as invalid register number to
13129 indirect_thunk_name and output_indirect_thunk.
13131 2018-02-02 Julia Koval <julia.koval@intel.com>
13133 * config.gcc: Add -march=icelake.
13134 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
13135 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
13136 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
13137 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
13138 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
13139 (processor_target_table): Add icelake.
13140 (ix86_option_override_internal): Handle new PTAs.
13141 (get_builtin_code_for_version): Handle icelake.
13142 (M_INTEL_COREI7_ICELAKE): New.
13143 (fold_builtin_cpu): Handle icelake.
13144 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
13145 * doc/invoke.texi: Add -march=icelake.
13147 2018-02-02 Julia Koval <julia.koval@intel.com>
13149 * config/i386/i386.c (ix86_option_override_internal): Change flags type
13150 to wide_int_bitmask.
13151 * wide-int-bitmask.h: New.
13153 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13156 * config/i386/i386.md: Replace Pmode with word_mode in
13157 builtin_setjmp_setup and builtin_longjmp to support x32.
13159 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
13163 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
13165 (rs6000_supported_cpu_names): New static variable.
13166 (linux_cpu_translation_table): Likewise.
13167 (elf_platform) <cpu>: Define new static variable and use it.
13168 Translate kernel AT_PLATFORM name to canonical name if needed.
13169 Error if platform name is unknown.
13171 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
13174 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
13176 2018-02-01 Jeff Law <law@redhat.com>
13179 * config/i386/i386.c (release_scratch_register_on_entry): Add new
13180 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
13181 the scratch if RELEASE_VIA_POP is false.
13182 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
13183 If we have to save a temporary register, decrement SIZE appropriately.
13184 Pass new arguments to release_scratch_register_on_entry.
13185 (ix86_adjust_stack_and_probe): Likewise.
13186 (ix86_emit_probe_stack_range): Pass new arguments to
13187 release_scratch_register_on_entry.
13189 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
13191 PR rtl-optimization/84157
13192 * combine.c (change_zero_ext): Use REG_P predicate in
13193 front of HARD_REGISTER_P predicate.
13195 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
13197 * config/avr/avr.c (avr_option_override): Move disabling of
13198 -fdelete-null-pointer-checks to...
13199 * common/config/avr/avr-common.c (avr_option_optimization_table):
13202 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13204 PR tree-optimization/81635
13205 * tree-data-ref.c (split_constant_offset_1): For types that
13206 wrap on overflow, try to use range info to prove that wrapping
13209 2018-02-01 Renlin Li <renlin.li@arm.com>
13212 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13213 TAILCALL_ADDR_REGS.
13214 (aarch64_register_move_cost): Likewise.
13215 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
13216 TAILCALL_ADDR_REGS.
13217 (REG_CLASS_NAMES): Likewise.
13218 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
13219 TAILCALL_ADDR_REGS. Remove IP registers.
13220 * config/aarch64/aarch64.md (Ucs): Update register constraint.
13222 2018-02-01 Richard Biener <rguenther@suse.de>
13224 * domwalk.h (dom_walker::dom_walker): Add additional constructor
13225 for specifying RPO order and allow NULL for that.
13226 * domwalk.c (dom_walker::dom_walker): Likewise.
13227 (dom_walker::walk): Handle NULL RPO order.
13228 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
13230 (rewrite_update_dom_walker): Likewise.
13231 (mark_def_dom_walker): Likewise.
13233 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13235 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
13236 (aarch64_maybe_expand_sve_subreg_move): Declare.
13237 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
13238 * config/aarch64/predicates.md (aarch64_any_register_operand): New
13240 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
13241 that are semantically a reverse operation.
13242 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
13243 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
13244 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
13246 (aarch64_can_change_mode_class): For big-endian, forbid changes
13247 between two SVE modes if they have different element sizes.
13249 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13251 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
13252 the TImode handling for big-endian targets.
13254 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13256 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
13257 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
13259 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
13260 Remove BSWAP handing for big-endian targets and use the form of
13261 LD1RQ appropariate for the mode.
13263 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13265 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
13266 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
13267 duplicated element.
13269 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
13272 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
13273 check for operands that need to go through aarch64_sve_reload_be.
13275 2018-02-01 Jakub Jelinek <jakub@redhat.com>
13277 PR tree-optimization/81661
13278 PR tree-optimization/84117
13279 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
13280 * tree-eh.c: Include gimplify.h.
13281 (find_trapping_overflow, replace_trapping_overflow,
13282 rewrite_to_non_trapping_overflow): New functions.
13283 * tree-vect-loop.c: Include tree-eh.h.
13284 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
13285 * tree-data-ref.c: Include tree-eh.h.
13286 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
13288 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
13290 PR rtl-optimization/84123
13291 * combine.c (change_zero_ext): Check if hard register satisfies
13292 can_change_dest_mode before calling gen_lowpart_SUBREG.
13294 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
13297 * ira.c (ira_init_register_move_cost): Remove assert.
13299 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13301 PR rtl-optimization/84071
13302 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
13303 * doc/tm.texi: Regenerate.
13305 2018-01-31 Richard Biener <rguenther@suse.de>
13307 PR tree-optimization/84132
13308 * tree-data-ref.c (analyze_miv_subscript): Properly
13309 check whether evolution_function_is_affine_multivariate_p
13310 before calling gcd_of_steps_may_divide_p.
13312 2018-01-31 Julia Koval <julia.koval@intel.com>
13315 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
13316 * config/i386/i386.md (rdpid_rex64) New.
13317 (rdpid): Make 32bit only.
13319 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
13322 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
13323 an IDENTIFIER_NODE for FUNCTION_TYPE's.
13325 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13328 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13330 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13332 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
13334 PR rtl-optimization/84071
13335 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
13336 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
13338 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13340 * config/arc/arc.c (arc_handle_aux_attribute): New function.
13341 (arc_attribute_table): Add 'aux' attribute.
13342 (arc_in_small_data_p): Consider aux like variables.
13343 (arc_is_aux_reg_p): New function.
13344 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
13345 (arc_get_aux_arg): New function.
13346 (prepare_move_operands): Handle aux-register access.
13347 (arc_handle_aux_attribute): New function.
13348 * doc/extend.texi (ARC Variable attributes): Add subsection.
13350 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
13352 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
13353 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
13354 (arc_attribute_table): Add 'uncached' attribute.
13355 (arc_print_operand): Print '.di' flag for uncached memory
13357 (arc_in_small_data_p): Do not consider for small data the uncached
13359 (arc_is_uncached_mem_p): New function.
13360 * config/arc/predicates.md (compact_store_memory_operand): Check
13361 for uncached memory accesses.
13362 (nonvol_nonimm_operand): Likewise.
13363 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
13365 2018-01-31 Jakub Jelinek <jakub@redhat.com>
13368 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
13369 falign-loops=): Add Optimization flag.
13371 2018-01-30 Jeff Law <law@redhat.com>
13374 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
13375 INT_REGISTERS_SAVED. Check it prior to calling
13376 get_scratch_register_on_entry.
13377 (ix86_adjust_stack_and_probe): Similarly.
13378 (ix86_emit_probe_stack_range): Similarly.
13379 (ix86_expand_prologue): Corresponding changes.
13381 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13384 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
13385 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
13387 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
13390 * lra-constraints.c (curr_insn_transform): Process AND in the
13393 2018-01-30 Jakub Jelinek <jakub@redhat.com>
13395 PR rtl-optimization/83986
13396 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
13397 dependence against last_pending_memory_flush in addition to
13398 pending_jump_insns.
13400 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
13402 PR tree-optimization/81611
13403 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
13406 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13409 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
13412 2018-01-30 Richard Biener <rguenther@suse.de>
13413 Jakub Jelinek <jakub@redhat.com>
13415 PR tree-optimization/84111
13416 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
13417 inner loops added during recursion, as they don't have up-to-date
13420 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13423 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
13424 (can_inline_edge_by_limits_p): ... here.
13425 (can_early_inline_edge_p, check_callers,
13426 update_caller_keys, update_callee_keys, recursive_inlining,
13427 add_new_edges_to_heap, speculation_useful_p,
13428 inline_small_functions,
13429 inline_small_functions, flatten_function,
13430 inline_to_all_callers_1): Update.
13432 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
13434 * profile-count.c (profile_count::combine_with_ipa_count): Handle
13437 2018-01-30 Richard Biener <rguenther@suse.de>
13439 PR tree-optimization/83008
13440 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
13441 invariant and constant vector uses in stmts when they need
13442 more than one stmt.
13444 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13447 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
13448 * configure: Regenerate.
13450 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13452 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
13454 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
13455 Use gen_rtx_REG rather than gen_lowpart.
13457 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13459 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
13460 rather than 0 when creating partial subregs.
13462 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
13464 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
13467 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
13470 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
13471 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
13472 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
13473 flags. This restores the settings used before the 2017-07-24.
13474 Turning off pre increment/decrement/modify allows IVOPTS to
13475 optimize DF/SF loops where the index is an int.
13477 2018-01-29 Richard Biener <rguenther@suse.de>
13478 Kelvin Nilsen <kelvin@gcc.gnu.org>
13481 * tree-vect-stmts.c (vectorizable_call): Don't call
13482 targetm.vectorize_builtin_md_vectorized_function if callee is
13485 2018-01-22 Carl Love <cel@us.ibm.com>
13487 * doc/extend.tex: Fix typo in second arg in
13488 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
13490 2018-01-29 Richard Biener <rguenther@suse.de>
13492 PR tree-optimization/84086
13493 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
13494 (flush_ssaname_freelist): When SSA names were released reset
13495 the SCEV hash table.
13497 2018-01-29 Richard Biener <rguenther@suse.de>
13499 PR tree-optimization/84057
13500 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
13501 removed paths when removing edges.
13503 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
13505 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
13506 -mfunction-return=@var{choice}.
13508 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
13510 PR diagnostic/84034
13511 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
13512 Handle CR like TAB.
13513 (layout::print_source_line): Likewise.
13514 (test_get_line_width_without_trailing_whitespace): Add test cases.
13516 2018-01-27 Jakub Jelinek <jakub@redhat.com>
13518 PR middle-end/84040
13519 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
13522 2018-01-26 Jim Wilson <jimw@sifive.com>
13524 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
13526 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
13529 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13531 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
13532 and CMP + SUB-immediate -> SUBS.
13534 2018-01-26 Martin Sebor <msebor@redhat.com>
13536 PR tree-optimization/83896
13537 * tree-ssa-strlen.c (get_string_len): Rename...
13538 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
13539 Avoid assuming length is constant.
13540 (handle_char_store): Use HOST_WIDE_INT for string length.
13542 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
13545 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
13546 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
13548 2018-01-26 Richard Biener <rguenther@suse.de>
13550 PR rtl-optimization/84003
13551 * dse.c (record_store): Only record redundant stores when
13552 the earlier store aliases at least all accesses the later one does.
13554 2018-01-26 Jakub Jelinek <jakub@redhat.com>
13556 PR rtl-optimization/83985
13557 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
13558 REG_CFA_RESTORE insns.
13559 (delete_unmarked_insns): Don't ignore separate shrink wrapping
13560 REG_CFA_RESTORE insns here.
13563 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
13564 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
13566 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13568 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
13569 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
13570 (arc_init): Likewise.
13571 (arc_override_options): Likewise.
13572 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
13574 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
13576 * config/arc/arc.h (TARGET_DBNZ): Define.
13577 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
13578 properly set the tune attribute.
13579 (dbnz): Use TARGET_DBNZ guard.
13580 * config/arc/arc.opt (mtune): Add core3 option.
13582 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13584 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
13585 recognize new pic like addresses.
13586 (arc_delegitimize_address): Clean up.
13588 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13590 * config/arc/arc-arches.def: Option mrf16 valid for all
13592 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
13593 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
13594 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
13595 * config/arc/arc-tables.opt: Regenerate.
13596 * config/arc/arc.c (arc_conditional_register_usage): Handle
13597 reduced register file case.
13598 (arc_file_start): Set must have build attributes.
13599 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
13600 mrf16 option value.
13601 * config/arc/arc.opt (mrf16): Add new option.
13602 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
13603 * config/arc/genmultilib.awk: Handle new mrf16 option.
13604 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
13605 * config/arc/t-multilib: Regenerate.
13606 * doc/invoke.texi (ARC Options): Document mrf16 option.
13608 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13610 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
13611 * config/arc/arc.c (arc_handle_secure_attribute): New function.
13612 (arc_attribute_table): Add 'secure_call' attribute.
13613 (arc_print_operand): Print secure call operand.
13614 (arc_function_ok_for_sibcall): Don't optimize tail calls when
13616 (arc_is_secure_call_p): New function. * config/arc/arc.md
13617 (call_i): Add support for sjli instruction.
13618 (call_value_i): Likewise.
13619 * config/arc/constraints.md (Csc): New constraint.
13621 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
13622 John Eric Martin <John.Martin@emmicro-us.com>
13624 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
13625 * config/arc/arc.c (_arc_jli_section): New struct.
13626 (arc_jli_section): New type.
13627 (rc_jli_sections): New static variable.
13628 (arc_handle_jli_attribute): New function.
13629 (arc_attribute_table): Add jli_always and jli_fixed attribute.
13630 (arc_file_end): New function.
13631 (TARGET_ASM_FILE_END): Define.
13632 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
13633 (arc_add_jli_section): New function.
13634 (jli_call_scan): Likewise.
13635 (arc_reorg): Call jli_call_scan.
13636 (arc_output_addsi): Remove 'S' from printing asm operand.
13637 (arc_is_jli_call_p): New function.
13638 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
13640 (movhi_insn): Likewise.
13641 (movsi_insn): Likewise.
13642 (movsi_set_cc_insn): Likewise.
13643 (loadqi_update): Likewise.
13644 (load_zeroextendqisi_update): Likewise.
13645 (load_signextendqisi_update): Likewise.
13646 (loadhi_update): Likewise.
13647 (load_zeroextendhisi_update): Likewise.
13648 (load_signextendhisi_update): Likewise.
13649 (loadsi_update): Likewise.
13650 (loadsf_update): Likewise.
13651 (movsicc_insn): Likewise.
13652 (bset_insn): Likewise.
13653 (bxor_insn): Likewise.
13654 (bclr_insn): Likewise.
13655 (bmsk_insn): Likewise.
13656 (bicsi3_insn): Likewise.
13657 (cmpsi_cc_c_insn): Likewise.
13658 (movsi_ne): Likewise.
13659 (movsi_cond_exec): Likewise.
13660 (clrsbsi2): Likewise.
13661 (norm_f): Likewise.
13671 (call_i): Remove 'S' asm letter, add jli instruction.
13672 (call_value_i): Likewise.
13673 * config/arc/arc.op (mjli-always): New option.
13674 * config/arc/constraints.md (Cji): New constraint.
13675 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
13677 (subsf3_fpx): Likewise.
13678 (mulsf3_fpx): Likewise.
13679 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
13681 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
13682 function attrbutes.
13683 * doc/invoke.texi (ARC): Document mjli-always option.
13685 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
13687 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
13688 avoid addition with 0 and use incw and decw where possible.
13690 2018-01-26 Richard Biener <rguenther@suse.de>
13692 PR tree-optimization/81082
13693 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
13694 association if it requires casting to unsigned.
13695 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
13696 from fold_plusminus_mult_expr to catch important cases late when
13697 range info is available.
13699 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13701 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
13702 * configure.ac (hidden_linkonce): New test.
13703 * configure: Regenerate.
13704 * config.in: Regenerate.
13706 2018-01-26 Julia Koval <julia.koval@intel.com>
13708 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
13709 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
13710 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
13711 _mm_mask_bitshuffle_epi64_mask): Fix type.
13712 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
13713 USI_FTYPE_V4DI_V4DI_USI): Remove.
13714 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
13715 __builtin_ia32_vpshufbitqmb256_mask,
13716 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
13717 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
13718 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
13720 2018-01-26 Alan Modra <amodra@gmail.com>
13723 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
13724 UNSPEC_VBPERMQ. Sort other unspecs.
13726 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
13728 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
13730 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
13732 PR middle-end/83055
13733 * predict.c (drop_profile): Do not push/pop cfun; update also
13735 (handle_missing_profiles): Fix logic looking for zero profiles.
13737 2018-01-25 Jakub Jelinek <jakub@redhat.com>
13739 PR middle-end/83977
13740 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
13741 on functions with #pragma omp declare simd or functions with simd
13743 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
13744 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
13745 Remove trailing \n from warning_at calls.
13747 2018-01-25 Tom de Vries <tom@codesourcery.com>
13750 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13751 for neutered workers.
13753 2018-01-24 Joseph Myers <joseph@codesourcery.com>
13756 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
13757 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
13759 2018-01-24 Jeff Law <law@redhat.com>
13762 * i386.c (get_probe_interval): Move to earlier point.
13763 (ix86_compute_frame_layout): If -fstack-clash-protection and
13764 the frame is larger than the probe interval, then use pushes
13765 to save registers rather than reg->mem moves.
13766 (ix86_expand_prologue): Remove conditional for int_registers_saved
13769 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
13772 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
13773 min/max for never referenced object.
13775 2018-01-24 Jakub Jelinek <jakub@redhat.com>
13777 PR middle-end/83977
13778 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
13780 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
13781 attributes from DECL_ATTRIBUTES (decl) without affecting
13782 DECL_ATTRIBUTES (current_function_decl).
13783 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
13784 functions with non-NULL DECL_ABSTRACT_ORIGIN.
13786 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
13788 PR tree-optimization/83979
13789 * fold-const.c (fold_comparison): Use constant_boolean_node
13790 instead of boolean_{true,false}_node.
13792 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
13794 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
13797 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13799 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
13800 Simplify the clause that sets the length attribute.
13801 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
13802 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
13803 clause that sets the length attribute.
13804 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
13806 2018-01-24 Tom de Vries <tom@codesourcery.com>
13809 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
13810 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
13811 Add strict parameter.
13812 (prevent_branch_around_nothing): Insert dummy insn between branch to
13813 label and label with no ptx insn inbetween.
13814 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
13816 2018-01-24 Tom de Vries <tom@codesourcery.com>
13819 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
13820 for neutered threads in warp.
13821 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
13823 2018-01-24 Richard Biener <rguenther@suse.de>
13825 PR tree-optimization/83176
13826 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
13829 2018-01-24 Richard Biener <rguenther@suse.de>
13831 PR tree-optimization/82819
13832 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
13833 code generating pluses that are no-ops in the target precision.
13835 2018-01-24 Richard Biener <rguenther@suse.de>
13837 PR middle-end/84000
13838 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
13840 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13842 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
13843 to merge probabilities.
13844 * predict.c (probably_never_executed): Also mark as cold functions
13845 with global 0 profile and guessed local profile.
13846 * profile-count.c (profile_probability::combine_with_count): New
13848 * profile-count.h (profile_probability::operator*,
13849 profile_probability::operator*=, profile_probability::operator/,
13850 profile_probability::operator/=): Reduce precision to adjusted
13851 and set value to guessed on contradictory divisions.
13852 (profile_probability::combine_with_freq): Remove.
13853 (profile_probability::combine_wiht_count): Declare.
13854 (profile_count::force_nonzero):: Set to adjusted.
13855 (profile_count::probability_in):: Set quality to adjusted.
13856 * tree-ssa-tail-merge.c (replace_block_by): Use
13857 combine_with_count.
13859 2018-01-23 Andrew Waterman <andrew@sifive.com>
13860 Jim Wilson <jimw@sifive.com>
13862 * config/riscv/riscv.c (riscv_stack_boundary): New.
13863 (riscv_option_override): Set riscv_stack_boundary. Handle
13864 riscv_preferred_stack_boundary_arg.
13865 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
13866 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
13867 (STACK_BOUNDARY): Set to riscv_stack_boundary.
13868 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
13869 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
13870 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
13872 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
13875 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
13876 of struct ix86_frame.
13877 (ix86_expand_epilogue): Likewise. Add a local variable for
13878 the reg_save_offset field in struct ix86_frame.
13880 2018-01-23 Bin Cheng <bin.cheng@arm.com>
13882 PR tree-optimization/82604
13883 * tree-loop-distribution.c (enum partition_kind): New enum item
13884 PKIND_PARTIAL_MEMSET.
13885 (partition_builtin_p): Support above new enum item.
13886 (generate_code_for_partition): Ditto.
13887 (compute_access_range): Differentiate cases that equality can be
13888 proven at all loops, the innermost loops or no loops.
13889 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
13890 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
13891 (finalize_partitions, distribute_loop): Don't fuse partition of
13892 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
13893 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
13894 parloop is enabled.
13896 2018-01-23 Martin Liska <mliska@suse.cz>
13898 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
13899 order to ignore the predictor.
13900 (PRED_POLYMORPHIC_CALL): Likewise.
13901 (PRED_RECURSIVE_CALL): Likewise.
13903 2018-01-23 Martin Liska <mliska@suse.cz>
13905 * tree-profile.c (tree_profiling): Print function header to
13906 aware reader which function we are working on.
13907 * value-prof.c (gimple_find_values_to_profile): Do not print
13908 not interesting value histograms.
13910 2018-01-23 Martin Liska <mliska@suse.cz>
13912 * profile-count.h (enum profile_quality): Add
13913 profile_uninitialized as the first value. Do not number values
13914 as they are zero based.
13915 (profile_count::verify): Update sanity check.
13916 (profile_probability::verify): Likewise.
13918 2018-01-23 Nathan Sidwell <nathan@acm.org>
13920 * doc/invoke.texi (ffor-scope): Deprecate.
13922 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13924 PR tree-optimization/83510
13925 * domwalk.c (set_all_edges_as_executable): New function.
13926 (dom_walker::dom_walker): Convert bool param
13927 "skip_unreachable_blocks" to enum reachability. Move setup of
13928 edge flags to set_all_edges_as_executable and only do it when
13929 reachability is REACHABLE_BLOCKS.
13930 * domwalk.h (enum dom_walker::reachability): New enum.
13931 (dom_walker::dom_walker): Convert bool param
13932 "skip_unreachable_blocks" to enum reachability.
13933 (set_all_edges_as_executable): New decl.
13934 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
13935 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
13937 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
13938 but converting true to REACHABLE_BLOCKS.
13939 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
13941 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
13942 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
13943 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
13945 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
13946 if check_all_array_refs will be called.
13948 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13950 * tree.c (selftest::test_location_wrappers): Add more test
13953 2018-01-23 David Malcolm <dmalcolm@redhat.com>
13955 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
13956 (selftest::test_bit_in_range): Likewise.
13958 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13961 * doc/sourcebuild.texi (vect_float): Say that the selector
13962 only describes the situation when -funsafe-math-optimizations is on.
13963 (vect_float_strict): Document.
13965 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
13967 PR tree-optimization/83965
13968 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13969 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
13970 instead of checking only for a reduction.
13971 (vect_recog_widen_sum_pattern): Likewise.
13973 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13975 * predict.c (probably_never_executed): Only use precise profile info.
13976 (compute_function_frequency): Skip after inlining hack since we now
13977 have quality checking.
13979 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
13981 * profile-count.h (profile_probability::very_unlikely,
13982 profile_probability::unlikely, profile_probability::even): Set
13983 precision to guessed.
13985 2018-01-23 Richard Biener <rguenther@suse.de>
13987 PR tree-optimization/83963
13988 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
13989 Properly terminate dominator walk when crossing the exit edge not
13990 when visiting its source block.
13992 2018-01-23 Jakub Jelinek <jakub@redhat.com>
13995 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
13996 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
13998 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14000 PR tree-optimization/83957
14001 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14002 semicolon after for body surrounded by braces.
14004 PR tree-optimization/83081
14005 * profile-count.h (profile_probability::split): New method.
14006 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14007 Use profile_probability::split.
14008 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14009 when splitting a single conditional jump into 2.
14011 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14013 PR tree-optimization/69452
14014 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14017 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14019 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14020 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14021 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14023 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14025 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14027 * config/rl78/rl78.md (movdi): New define_expand.
14028 * config/rl78/rl78.c (rl78_split_movdi): New function.
14030 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14033 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14035 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14036 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14037 128-bit to produce an UNSPEC move to get the double word with the
14038 signbit and then a shift directly to do signbit.
14039 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14040 implementation with a new version that just does either a direct
14041 move or a regular move. Move memory interface to separate insns.
14042 Move insns so they are next to the expander.
14043 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14044 with signbit move. Split big and little endian case.
14045 (signbit<mode>2_dm_mem_le): Likewise.
14046 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14047 (signbit<mode>2_dm2): Likewise.
14049 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14051 * config/rl78/rl78.md (anddi3): New define_expand.
14053 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14055 * config/rl78/rl78.md (umindi3): New define_expand.
14057 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14059 * config/rl78/rl78.md (smindi3): New define_expand.
14061 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14063 * config/rl78/rl78.md (smaxdi3): New define_expand.
14065 2018-01-22 Carl Love <cel@us.ibm.com>
14067 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
14068 LVX_V1TI): Add macro expansion.
14069 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
14070 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
14071 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
14072 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
14073 Change check to determine if the instruction is a byte reversing
14074 entry. Fix typo in comment.
14075 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
14076 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
14077 Add def_builtin calls for new builtins.
14078 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
14079 Add define_insn expansion.
14081 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14083 * config/rl78/rl78.md (umaxdi3): New define_expand.
14085 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14087 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
14088 for non-QImode registers.
14090 2018-01-22 Richard Biener <rguenther@suse.de>
14092 PR tree-optimization/83963
14093 * graphite-scop-detection.c (scop_detection::get_sese): Delay
14094 including the loop exit block.
14095 (scop_detection::merge_sese): Likewise.
14096 (scop_detection::add_scop): Do it here instead.
14098 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14100 * doc/sourcebuild.texi (arm_softfloat): Document.
14102 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
14105 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
14106 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
14107 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
14109 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14110 David Edelsohn <dje.gcc@gmail.com>
14113 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14114 Change "crset eq" to "crset 2".
14115 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14116 (*call_indirect_aix<mode>_nospec): Likewise.
14117 (*call_value_indirect_aix<mode>_nospec): Likewise.
14118 (*call_indirect_elfv2<mode>_nospec): Likewise.
14119 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
14120 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
14121 change assembly output from . to $.
14122 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14123 (indirect_jump<mode>_nospec): Change assembly output from . to $.
14124 (*tablejump<mode>_internal1_nospec): Likewise.
14126 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
14129 * config/sh/sh_optimize_sett_clrt.cc:
14130 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
14132 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
14134 PR tree-optimization/83940
14135 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
14136 offset_dt to vect_constant_def rather than vect_unknown_def_type.
14137 (vect_check_load_store_mask): Add a mask_dt_out parameter and
14138 use it to pass back the definition type.
14139 (vect_check_store_rhs): Likewise rhs_dt_out.
14140 (vect_build_gather_load_calls): Add a mask_dt argument and use
14141 it instead of a call to vect_is_simple_use.
14142 (vectorizable_store): Update calls to vect_check_load_store_mask
14143 and vect_check_store_rhs. Use the dt returned by the latter instead
14144 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
14145 instead of calls to vect_is_simple_use. Pass the scalar rather
14146 than the vector operand to vect_is_simple_use when handling
14147 second and subsequent copies of an rhs value.
14148 (vectorizable_load): Update calls to vect_check_load_store_mask
14149 and vect_build_gather_load_calls. Use the cached mask_dt and
14150 gs_info.offset_dt instead of calls to vect_is_simple_use.
14152 2018-01-20 Jakub Jelinek <jakub@redhat.com>
14154 PR middle-end/83945
14155 * tree-emutls.c: Include gimplify.h.
14156 (lower_emutls_2): New function.
14157 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
14158 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
14159 it before further processing.
14162 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
14163 UINTVAL (trueop1) instead of INTVAL (op1).
14165 2018-01-19 Jakub Jelinek <jakub@redhat.com>
14169 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
14170 INCOMING_FRAME_SP_OFFSET if not defined.
14171 (scan_trace): Add ENTRY argument. If true and
14172 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
14173 emit a note to adjust the CFA offset.
14174 (create_cfi_notes): Adjust scan_trace callers.
14175 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
14176 INCOMING_FRAME_SP_OFFSET in the CIE.
14177 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
14178 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
14180 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
14181 * doc/tm.texi: Regenerated.
14183 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14185 PR rtl-optimization/83147
14186 * lra-constraints.c (remove_inheritance_pseudos): Use
14187 lra_substitute_pseudo_within_insn.
14189 2018-01-19 Tom de Vries <tom@codesourcery.com>
14190 Cesar Philippidis <cesar@codesourcery.com>
14193 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
14195 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
14198 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
14199 spaces for function labels.
14201 2018-01-19 Martin Liska <mliska@suse.cz>
14203 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
14204 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
14205 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
14206 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
14207 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
14208 (PRED_CONST_RETURN): Change from 69 to 65.
14209 (PRED_NULL_RETURN): Change from 91 to 71.
14210 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
14211 (PRED_LOOP_GUARD): Change from 66 to 73.
14213 2018-01-19 Martin Liska <mliska@suse.cz>
14215 * predict.c (predict_insn_def): Add new assert.
14216 (struct branch_predictor): Change type to signed integer.
14217 (test_prediction_value_range): Amend test to cover
14218 PROB_UNINITIALIZED.
14219 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
14220 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
14221 (PRED_LOOP_ITERATIONS_MAX): Likewise.
14222 (PRED_LOOP_IV_COMPARE): Likewise.
14223 * predict.h (PROB_UNINITIALIZED): Define new constant.
14225 2018-01-19 Martin Liska <mliska@suse.cz>
14227 * predict.c (dump_prediction): Add new format for
14228 analyze_brprob.py script which is enabled with -details
14230 * profile-count.h (precise_p): New function.
14232 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14234 PR tree-optimization/83922
14235 * tree-vect-loop.c (vect_verify_full_masking): Return false if
14236 there are no statements that need masking.
14237 (vect_active_double_reduction_p): New function.
14238 (vect_analyze_loop_operations): Use it when handling phis that
14239 are not in the loop header.
14241 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
14243 PR tree-optimization/83914
14244 * tree-vect-loop.c (vectorizable_induction): Don't convert
14245 init_expr or apply the peeling adjustment for inductions
14246 that are nested within the vectorized loop.
14248 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14250 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
14253 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14257 * function.h (gimplify_parameters): Add gimple_seq * argument.
14258 * function.c: Include gimple.h and options.h.
14259 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
14260 for the added local temporaries if needed.
14261 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
14262 if there are any parameter cleanups, wrap whole body into a
14263 try/finally with the cleanups.
14265 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
14268 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
14269 Use GET_MODE_CLASS for scalar floating point.
14271 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14275 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
14276 Fix call of call_cgraph_insertion_hooks.
14278 2018-01-18 Martin Sebor <msebor@redhat.com>
14280 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
14282 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
14285 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
14288 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
14291 * common.opt: (-ffile-prefix-map): New option.
14292 * opts.c (common_handle_option): Defer it.
14293 * opts-global.c (handle_common_deferred_options): Handle it.
14294 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
14295 * file-prefix-map.h: New file.
14296 (remap_debug_filename, add_debug_prefix_map): ...here.
14297 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
14298 * final.c (debug_prefix_map, add_debug_prefix_map
14299 remap_debug_filename): Move to...
14300 * file-prefix-map.c: New file.
14301 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
14302 generalize, get rid of alloca(), use strrchr() instead of strchr().
14303 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
14304 Implement in terms of add_prefix_map().
14305 (remap_macro_filename, remap_debug_filename): Implement in term of
14307 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
14308 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
14309 * dbxout.c: Include file-prefix-map.h.
14310 * varasm.c: Likewise.
14311 * vmsdbgout.c: Likewise.
14312 * xcoffout.c: Likewise.
14313 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
14314 * doc/cppopts.texi (-fmacro-prefix-map): Document.
14315 * doc/invoke.texi (-ffile-prefix-map): Document.
14316 (-fdebug-prefix-map): Update description.
14318 2018-01-18 Martin Liska <mliska@suse.cz>
14320 * config/i386/i386.c (indirect_thunk_name): Document that also
14322 (output_indirect_thunk): Document why both instructions
14323 (pause and lfence) are generated.
14325 2018-01-18 Richard Biener <rguenther@suse.de>
14327 PR tree-optimization/83887
14328 * graphite-scop-detection.c
14329 (scop_detection::get_nearest_dom_with_single_entry): Remove.
14330 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
14331 (scop_detection::merge_sese): Re-implement with a flood-fill
14332 algorithm that properly finds a SESE region if it exists.
14334 2018-01-18 Jakub Jelinek <jakub@redhat.com>
14337 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
14338 pointer_diff optimizations use view_convert instead of convert.
14340 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14342 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14343 Generate different code for -mno-speculate-indirect-jumps.
14344 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14345 (*call_indirect_aix<mode>): Disable for
14346 -mno-speculate-indirect-jumps.
14347 (*call_indirect_aix<mode>_nospec): New define_insn.
14348 (*call_value_indirect_aix<mode>): Disable for
14349 -mno-speculate-indirect-jumps.
14350 (*call_value_indirect_aix<mode>_nospec): New define_insn.
14351 (*sibcall_nonlocal_sysv<mode>): Generate different code for
14352 -mno-speculate-indirect-jumps.
14353 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14355 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14357 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
14358 long double type, set the flags for noting the default long double
14359 type, even if we don't pass or return a long double type.
14361 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
14364 * ipa-inline.c (flatten_function): Do not overwrite final inlining
14367 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
14369 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
14370 support for merge[hl].
14371 (fold_mergehl_helper): New helper function.
14372 (tree-vector-builder.h): New #include for tree_vector_builder usage.
14373 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
14374 (altivec_vmrglw_direct): Add xxmrglw insn.
14376 2018-01-17 Andrew Waterman <andrew@sifive.com>
14378 * config/riscv/riscv.c (riscv_conditional_register_usage): If
14379 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
14381 2018-01-17 David Malcolm <dmalcolm@redhat.com>
14384 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
14385 call the lto_location_cache before reading the
14386 DECL_SOURCE_LOCATION of the types.
14388 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
14389 Richard Sandiford <richard.sandiford@linaro.org>
14391 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
14392 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
14393 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
14394 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
14395 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
14397 * config/aarch64/constraints.md (aarch64_movti_operand):
14399 * config/aarch64/predicates.md (Uti): Add new constraint.
14401 2018-01-17 Carl Love <cel@us.ibm.com>
14403 * config/rs6000/vsx.md (define_expand xl_len_r,
14404 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
14405 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
14407 (define_expand, define_insn): Move the shift left from the
14408 define_insn to the define_expand for lxvl and stxvl instructions.
14409 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
14410 and XL_LEN_R definitions to PURE.
14412 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
14414 * config/i386/i386.c (indirect_thunk_name): Declare regno
14415 as unsigned int. Compare regno with INVALID_REGNUM.
14416 (output_indirect_thunk): Ditto.
14417 (output_indirect_thunk_function): Ditto.
14418 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
14419 in the call to output_indirect_thunk_function.
14421 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
14423 PR middle-end/83884
14424 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
14425 rather than the size of inner_type to determine the stack slot size
14426 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
14428 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
14431 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
14434 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
14436 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
14437 endian Linux systems to optionally enable multilibs for selecting
14438 the long double type if the user configured an explicit type.
14439 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
14440 have no long double multilibs if not defined.
14441 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
14442 warn if the user used -mabi={ieee,ibm}longdouble and we built
14443 multilibs for long double.
14444 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
14445 appropriate multilib option.
14446 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
14448 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
14449 for building long double multilibs.
14450 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
14452 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
14454 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
14457 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
14459 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
14462 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
14465 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
14468 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14470 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
14471 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
14473 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14475 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
14476 different rtl trees depending on TARGET_64BIT.
14477 (rs6000_gen_lvx): Likewise.
14479 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
14481 * config/visium/visium.md (nop): Tweak comment.
14482 (hazard_nop): Likewise.
14484 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14486 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
14487 -mspeculate-indirect-jumps.
14488 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
14489 for -mno-speculate-indirect-jumps.
14490 (*call_indirect_elfv2<mode>_nospec): New define_insn.
14491 (*call_value_indirect_elfv2<mode>): Disable for
14492 -mno-speculate-indirect-jumps.
14493 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
14494 (indirect_jump): Emit different RTL for
14495 -mno-speculate-indirect-jumps.
14496 (*indirect_jump<mode>): Disable for
14497 -mno-speculate-indirect-jumps.
14498 (*indirect_jump<mode>_nospec): New define_insn.
14499 (tablejump): Emit different RTL for
14500 -mno-speculate-indirect-jumps.
14501 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
14502 (tablejumpsi_nospec): New define_expand.
14503 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
14504 (tablejumpdi_nospec): New define_expand.
14505 (*tablejump<mode>_internal1): Disable for
14506 -mno-speculate-indirect-jumps.
14507 (*tablejump<mode>_internal1_nospec): New define_insn.
14508 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
14511 2018-01-16 Artyom Skrobov tyomitch@gmail.com
14513 * caller-save.c (insert_save): Drop unnecessary parameter. All
14516 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14517 Richard Biener <rguenth@suse.de>
14520 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
14521 return early, inline manually is_gimple_sizepos. Make sure if we
14522 call gimplify_expr we don't end up with a gimple constant.
14523 * tree.c (variably_modified_type_p): Don't return true for
14524 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
14525 * gimplify.h (is_gimple_sizepos): Remove.
14527 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14529 PR tree-optimization/83857
14530 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
14531 vectorizable_live_operation for pure SLP statements.
14532 (vectorizable_live_operation): Handle PHIs.
14534 2018-01-16 Richard Biener <rguenther@suse.de>
14536 PR tree-optimization/83867
14537 * tree-vect-stmts.c (vect_transform_stmt): Precompute
14538 nested_in_vect_loop_p since the scalar stmt may get invalidated.
14540 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14543 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
14544 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
14545 If off is not INTEGER_CST, issue a may not be aligned warning
14546 rather than isn't aligned. Use isn%'t rather than isn't.
14547 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
14549 <case MULT_EXPR>: Improve the case when bottom and one of the
14550 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
14551 operand, in that case check if the other operand is multiple of
14552 bottom divided by the INTEGER_CST operand.
14554 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14557 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
14558 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
14559 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
14560 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
14561 * config/pa/pa.c (pa_function_arg_advance): Likewise.
14562 (pa_function_arg, pa_arg_partial_bytes): Likewise.
14563 (pa_function_arg_size): New function.
14565 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14567 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
14568 in a separate statement.
14570 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
14572 PR tree-optimization/83847
14573 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
14574 group gathers and scatters.
14576 2018-01-16 Jakub Jelinek <jakub@redhat.com>
14578 PR rtl-optimization/86620
14579 * params.def (max-sched-ready-insns): Bump minimum value to 1.
14581 PR rtl-optimization/83213
14582 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
14583 to last if both are JUMP_INSNs.
14585 PR tree-optimization/83843
14586 * gimple-ssa-store-merging.c
14587 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
14588 store_immediate_info for bswap/nop orig_stores.
14590 2018-01-15 Andrew Waterman <andrew@sifive.com>
14592 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
14594 <UDIV>: Increase cost if !TARGET_DIV.
14596 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
14598 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
14599 (define_attr "cr_logical_3op"): New.
14600 (cceq_ior_compare): Adjust.
14601 (cceq_ior_compare_complement): Adjust.
14602 (*cceq_rev_compare): Adjust.
14603 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
14604 (is_cracked_insn): Adjust.
14605 (insn_must_be_first_in_group): Adjust.
14606 * config/rs6000/40x.md: Adjust.
14607 * config/rs6000/440.md: Adjust.
14608 * config/rs6000/476.md: Adjust.
14609 * config/rs6000/601.md: Adjust.
14610 * config/rs6000/603.md: Adjust.
14611 * config/rs6000/6xx.md: Adjust.
14612 * config/rs6000/7450.md: Adjust.
14613 * config/rs6000/7xx.md: Adjust.
14614 * config/rs6000/8540.md: Adjust.
14615 * config/rs6000/cell.md: Adjust.
14616 * config/rs6000/e300c2c3.md: Adjust.
14617 * config/rs6000/e500mc.md: Adjust.
14618 * config/rs6000/e500mc64.md: Adjust.
14619 * config/rs6000/e5500.md: Adjust.
14620 * config/rs6000/e6500.md: Adjust.
14621 * config/rs6000/mpc.md: Adjust.
14622 * config/rs6000/power4.md: Adjust.
14623 * config/rs6000/power5.md: Adjust.
14624 * config/rs6000/power6.md: Adjust.
14625 * config/rs6000/power7.md: Adjust.
14626 * config/rs6000/power8.md: Adjust.
14627 * config/rs6000/power9.md: Adjust.
14628 * config/rs6000/rs64.md: Adjust.
14629 * config/rs6000/titan.md: Adjust.
14631 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14633 * config/i386/predicates.md (indirect_branch_operand): Rewrite
14634 ix86_indirect_branch_register logic.
14636 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14638 * config/i386/constraints.md (Bs): Update
14639 ix86_indirect_branch_register check. Don't check
14640 ix86_indirect_branch_register with GOT_memory_operand.
14642 * config/i386/predicates.md (GOT_memory_operand): Don't check
14643 ix86_indirect_branch_register here.
14644 (GOT32_symbol_operand): Likewise.
14646 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14648 * config/i386/predicates.md (constant_call_address_operand):
14649 Rewrite ix86_indirect_branch_register logic.
14650 (sibcall_insn_operand): Likewise.
14652 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14654 * config/i386/constraints.md (Bs): Replace
14655 ix86_indirect_branch_thunk_register with
14656 ix86_indirect_branch_register.
14658 * config/i386/i386.md (indirect_jump): Likewise.
14659 (tablejump): Likewise.
14660 (*sibcall_memory): Likewise.
14661 (*sibcall_value_memory): Likewise.
14662 Peepholes of indirect call and jump via memory: Likewise.
14663 * config/i386/i386.opt: Likewise.
14664 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14665 (GOT_memory_operand): Likewise.
14666 (call_insn_operand): Likewise.
14667 (sibcall_insn_operand): Likewise.
14668 (GOT32_symbol_operand): Likewise.
14670 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14672 PR middle-end/83837
14673 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
14674 type rather than type addr's type points to.
14675 (expand_omp_atomic_mutex): Likewise.
14676 (expand_omp_atomic): Likewise.
14678 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
14681 * config/i386/i386.c (output_indirect_thunk_function): Use
14682 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
14683 for __x86_return_thunk.
14685 2018-01-15 Richard Biener <rguenther@suse.de>
14687 PR middle-end/83850
14688 * expmed.c (extract_bit_field_1): Fix typo.
14690 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14693 * config/arm/iterators.md (VF): New mode iterator.
14694 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
14695 Remove integer-related logic from pattern.
14696 (neon_vabd<mode>_3): Likewise.
14698 2018-01-15 Jakub Jelinek <jakub@redhat.com>
14700 PR middle-end/82694
14701 * common.opt (fstrict-overflow): No longer an alias.
14702 (fwrapv-pointer): New option.
14703 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
14704 also for pointer types based on flag_wrapv_pointer.
14705 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
14706 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
14707 opts->x_flag_wrapv got set.
14708 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
14709 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
14710 POINTER_TYPE_OVERFLOW_UNDEFINED.
14711 * match.pd: Likewise in address comparison pattern.
14712 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
14714 2018-01-15 Richard Biener <rguenther@suse.de>
14717 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
14718 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
14719 Reset type names to their identifier if their TYPE_DECL doesn't
14720 have linkage (and thus is used for ODR and devirt).
14721 (save_debug_info_for_decl): Remove.
14722 (save_debug_info_for_type): Likewise.
14723 (add_tree_to_fld_list): Adjust.
14724 * tree-pretty-print.c (dump_generic_node): Make dumping of
14725 type names more robust.
14727 2018-01-15 Richard Biener <rguenther@suse.de>
14729 * BASE-VER: Bump to 8.0.1.
14731 2018-01-14 Martin Sebor <msebor@redhat.com>
14734 * builtins.c (check_access): Avoid warning when the no-warning bit
14737 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
14739 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
14740 * ira-color (allocno_hard_regs_compare): Likewise.
14742 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
14745 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
14746 Use .pushsection/.popsection.
14748 2018-01-14 Martin Sebor <msebor@redhat.com>
14751 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
14753 2018-01-14 Jakub Jelinek <jakub@redhat.com>
14755 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
14756 entry from extra_headers.
14757 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
14758 extra_headers, make the list bitwise identical to the i?86-*-* one.
14760 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14762 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
14763 -mcmodel=large with -mindirect-branch=thunk,
14764 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
14765 -mfunction-return=thunk-extern.
14766 * doc/invoke.texi: Document -mcmodel=large is incompatible with
14767 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
14768 -mfunction-return=thunk and -mfunction-return=thunk-extern.
14770 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14772 * config/i386/i386.c (print_reg): Print the name of the full
14773 integer register without '%'.
14774 (ix86_print_operand): Handle 'V'.
14775 * doc/extend.texi: Document 'V' modifier.
14777 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14779 * config/i386/constraints.md (Bs): Disallow memory operand for
14780 -mindirect-branch-register.
14782 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14783 (GOT_memory_operand): Likewise.
14784 (call_insn_operand): Likewise.
14785 (sibcall_insn_operand): Likewise.
14786 (GOT32_symbol_operand): Likewise.
14787 * config/i386/i386.md (indirect_jump): Call convert_memory_address
14788 for -mindirect-branch-register.
14789 (tablejump): Likewise.
14790 (*sibcall_memory): Likewise.
14791 (*sibcall_value_memory): Likewise.
14792 Disallow peepholes of indirect call and jump via memory for
14793 -mindirect-branch-register.
14794 (*call_pop): Replace m with Bw.
14795 (*call_value_pop): Likewise.
14796 (*sibcall_pop_memory): Replace m with Bs.
14797 * config/i386/i386.opt (mindirect-branch-register): New option.
14798 * doc/invoke.texi: Document -mindirect-branch-register option.
14800 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14802 * config/i386/i386-protos.h (ix86_output_function_return): New.
14803 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
14804 set function_return_type.
14805 (indirect_thunk_name): Add ret_p to indicate thunk for function
14807 (output_indirect_thunk_function): Pass false to
14808 indirect_thunk_name.
14809 (ix86_output_indirect_branch_via_reg): Likewise.
14810 (ix86_output_indirect_branch_via_push): Likewise.
14811 (output_indirect_thunk_function): Create alias for function
14812 return thunk if regno < 0.
14813 (ix86_output_function_return): New function.
14814 (ix86_handle_fndecl_attribute): Handle function_return.
14815 (ix86_attribute_table): Add function_return.
14816 * config/i386/i386.h (machine_function): Add
14817 function_return_type.
14818 * config/i386/i386.md (simple_return_internal): Use
14819 ix86_output_function_return.
14820 (simple_return_internal_long): Likewise.
14821 * config/i386/i386.opt (mfunction-return=): New option.
14822 (indirect_branch): Mention -mfunction-return=.
14823 * doc/extend.texi: Document function_return function attribute.
14824 * doc/invoke.texi: Document -mfunction-return= option.
14826 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
14828 * config/i386/i386-opts.h (indirect_branch): New.
14829 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
14830 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
14831 with local indirect jump when converting indirect call and jump.
14832 (ix86_set_indirect_branch_type): New.
14833 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
14834 (indirectlabelno): New.
14835 (indirect_thunk_needed): Likewise.
14836 (indirect_thunk_bnd_needed): Likewise.
14837 (indirect_thunks_used): Likewise.
14838 (indirect_thunks_bnd_used): Likewise.
14839 (INDIRECT_LABEL): Likewise.
14840 (indirect_thunk_name): Likewise.
14841 (output_indirect_thunk): Likewise.
14842 (output_indirect_thunk_function): Likewise.
14843 (ix86_output_indirect_branch_via_reg): Likewise.
14844 (ix86_output_indirect_branch_via_push): Likewise.
14845 (ix86_output_indirect_branch): Likewise.
14846 (ix86_output_indirect_jmp): Likewise.
14847 (ix86_code_end): Call output_indirect_thunk_function if needed.
14848 (ix86_output_call_insn): Call ix86_output_indirect_branch if
14850 (ix86_handle_fndecl_attribute): Handle indirect_branch.
14851 (ix86_attribute_table): Add indirect_branch.
14852 * config/i386/i386.h (machine_function): Add indirect_branch_type
14853 and has_local_indirect_jump.
14854 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
14856 (tablejump): Likewise.
14857 (*indirect_jump): Use ix86_output_indirect_jmp.
14858 (*tablejump_1): Likewise.
14859 (simple_return_indirect_internal): Likewise.
14860 * config/i386/i386.opt (mindirect-branch=): New option.
14861 (indirect_branch): New.
14864 (thunk-inline): Likewise.
14865 (thunk-extern): Likewise.
14866 * doc/extend.texi: Document indirect_branch function attribute.
14867 * doc/invoke.texi: Document -mindirect-branch= option.
14869 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
14872 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
14874 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
14876 * ipa-inline.c (want_inline_small_function_p): Return false if
14877 inlining has already failed with CIF_FINAL_ERROR.
14878 (update_caller_keys): Call want_inline_small_function_p before
14880 (update_callee_keys): Likewise.
14882 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14884 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
14886 (rs6000_quadword_masked_address_p): Likewise.
14887 (quad_aligned_load_p): Likewise.
14888 (quad_aligned_store_p): Likewise.
14889 (const_load_sequence_p): Add comment to describe the outer-most loop.
14890 (mimic_memory_attributes_and_flags): New function.
14891 (rs6000_gen_stvx): Likewise.
14892 (replace_swapped_aligned_store): Likewise.
14893 (rs6000_gen_lvx): Likewise.
14894 (replace_swapped_aligned_load): Likewise.
14895 (replace_swapped_load_constant): Capitalize argument name in
14896 comment describing this function.
14897 (rs6000_analyze_swaps): Add a third pass to search for vector loads
14898 and stores that access quad-word aligned addresses and replace
14899 with stvx or lvx instructions when appropriate.
14900 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
14901 New function prototype.
14902 (rs6000_quadword_masked_address_p): Likewise.
14903 (rs6000_gen_lvx): Likewise.
14904 (rs6000_gen_stvx): Likewise.
14905 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
14906 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
14907 when memory address is aligned.
14908 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
14909 this split to select lvx instruction when memory address is aligned.
14910 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
14911 instruction when memory address is aligned.
14912 (*vsx_le_perm_load_v16qi): Likewise.
14913 (four unnamed splitters): Modify to select the stvx instruction
14914 when memory is aligned.
14916 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
14918 * predict.c (determine_unlikely_bbs): Handle correctly BBs
14919 which appears in the queue multiple times.
14921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14922 Alan Hayward <alan.hayward@arm.com>
14923 David Sherwood <david.sherwood@arm.com>
14925 * tree-vectorizer.h (vec_lower_bound): New structure.
14926 (_loop_vec_info): Add check_nonzero and lower_bounds.
14927 (LOOP_VINFO_CHECK_NONZERO): New macro.
14928 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
14929 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
14930 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
14931 fields. Make seg_len the distance travelled, not including the
14933 (dr_direction_indicator): Declare.
14934 (dr_zero_step_indicator): Likewise.
14935 (dr_known_forward_stride_p): Likewise.
14936 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
14938 (runtime_alias_check_p): Allow runtime alias checks with
14940 (operator ==): Compare access_size and align.
14941 (prune_runtime_alias_test_list): Rework for new distinction between
14942 the access_size and seg_len.
14943 (create_intersect_range_checks_index): Likewise. Cope with polynomial
14945 (get_segment_min_max): New function.
14946 (create_intersect_range_checks): Use it.
14947 (dr_step_indicator): New function.
14948 (dr_direction_indicator): Likewise.
14949 (dr_zero_step_indicator): Likewise.
14950 (dr_known_forward_stride_p): Likewise.
14951 * tree-loop-distribution.c (data_ref_segment_size): Return
14952 DR_STEP * (niters - 1).
14953 (compute_alias_check_pairs): Update call to the dr_with_seg_len
14955 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
14956 (vect_preserves_scalar_order_p): New function, split out from...
14957 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
14958 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
14959 (vect_vfa_access_size): New function.
14960 (vect_vfa_align): Likewise.
14961 (vect_compile_time_alias): Take access_size_a and access_b arguments.
14962 (dump_lower_bound): New function.
14963 (vect_check_lower_bound): Likewise.
14964 (vect_small_gap_p): Likewise.
14965 (vectorizable_with_step_bound_p): Likewise.
14966 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
14967 depencies if the vectorization factor is 1. Convert the checks
14968 for nonzero steps into checks on the bounds of DR_STEP. Try using
14969 a bunds check for variable steps if the minimum required step is
14970 relatively small. Update calls to the dr_with_seg_len
14971 constructor and to vect_compile_time_alias.
14972 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
14974 (vect_loop_versioning): Call it.
14975 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
14977 (vect_estimate_min_profitable_iters): Account for any bounds checks.
14979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14980 Alan Hayward <alan.hayward@arm.com>
14981 David Sherwood <david.sherwood@arm.com>
14983 * doc/sourcebuild.texi (vect_scatter_store): Document.
14984 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
14986 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
14988 * genopinit.c (main): Add supports_vec_scatter_store and
14989 supports_vec_scatter_store_cached to target_optabs.
14990 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
14991 IFN_MASK_SCATTER_STORE.
14992 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
14994 * internal-fn.h (internal_store_fn_p): Declare.
14995 (internal_fn_stored_value_index): Likewise.
14996 * internal-fn.c (scatter_store_direct): New macro.
14997 (expand_scatter_store_optab_fn): New function.
14998 (direct_scatter_store_optab_supported_p): New macro.
14999 (internal_store_fn_p): New function.
15000 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15001 IFN_MASK_SCATTER_STORE.
15002 (internal_fn_mask_index): Likewise.
15003 (internal_fn_stored_value_index): New function.
15004 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15005 for scatter stores.
15006 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15007 * optabs-query.c (supports_vec_scatter_store_p): New function.
15008 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15009 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15010 true for scatter stores.
15011 (vect_gather_scatter_fn_p): Handle scatter stores too.
15012 (vect_check_gather_scatter): Consider using scatter stores if
15013 supports_vec_scatter_store_p.
15014 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15015 scatter stores too.
15016 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15017 internal_fn_stored_value_index.
15018 (check_load_store_masking): Handle scatter stores too.
15019 (vect_get_store_rhs): Make public.
15020 (vectorizable_call): Use internal_store_fn_p.
15021 (vectorizable_store): Handle scatter store internal functions.
15022 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15023 when deciding whether the end of the group has been reached.
15024 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15025 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15026 (mask_scatter_store<mode>): New insns.
15028 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15029 Alan Hayward <alan.hayward@arm.com>
15030 David Sherwood <david.sherwood@arm.com>
15032 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15033 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15034 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15036 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15037 Use vect_truncate_gather_scatter_offset if we can't treat the
15038 operation as a normal gather load or scatter store.
15039 (get_group_load_store_type): Take the gather_scatter_info
15040 as argument. Try using a gather load or scatter store for
15041 single-element groups.
15042 (get_load_store_type): Update calls to get_group_load_store_type
15043 and vect_use_strided_gather_scatters_p.
15045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15046 Alan Hayward <alan.hayward@arm.com>
15047 David Sherwood <david.sherwood@arm.com>
15049 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15050 optional tree argument.
15051 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15053 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15054 but continue to use the current value as a fallback.
15055 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15056 to compare the updates.
15057 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15058 (get_load_store_type): Use it when handling a strided access.
15059 (vect_get_strided_load_store_ops): New function.
15060 (vect_get_data_ptr_increment): Likewise.
15061 (vectorizable_load): Handle strided gather loads. Always pass
15062 a step to vect_create_data_ref_ptr and bump_vector_ptr.
15064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15065 Alan Hayward <alan.hayward@arm.com>
15066 David Sherwood <david.sherwood@arm.com>
15068 * doc/md.texi (gather_load@var{m}): Document.
15069 (mask_gather_load@var{m}): Likewise.
15070 * genopinit.c (main): Add supports_vec_gather_load and
15071 supports_vec_gather_load_cached to target_optabs.
15072 * optabs-tree.c (init_tree_optimization_optabs): Use
15073 ggc_cleared_alloc to allocate target_optabs.
15074 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
15075 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
15077 * internal-fn.h (internal_load_fn_p): Declare.
15078 (internal_gather_scatter_fn_p): Likewise.
15079 (internal_fn_mask_index): Likewise.
15080 (internal_gather_scatter_fn_supported_p): Likewise.
15081 * internal-fn.c (gather_load_direct): New macro.
15082 (expand_gather_load_optab_fn): New function.
15083 (direct_gather_load_optab_supported_p): New macro.
15084 (direct_internal_fn_optab): New function.
15085 (internal_load_fn_p): Likewise.
15086 (internal_gather_scatter_fn_p): Likewise.
15087 (internal_fn_mask_index): Likewise.
15088 (internal_gather_scatter_fn_supported_p): Likewise.
15089 * optabs-query.c (supports_at_least_one_mode_p): New function.
15090 (supports_vec_gather_load_p): Likewise.
15091 * optabs-query.h (supports_vec_gather_load_p): Declare.
15092 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
15093 and memory_type field.
15094 (NUM_PATTERNS): Bump to 15.
15095 * tree-vect-data-refs.c: Include internal-fn.h.
15096 (vect_gather_scatter_fn_p): New function.
15097 (vect_describe_gather_scatter_call): Likewise.
15098 (vect_check_gather_scatter): Try using internal functions for
15099 gather loads. Recognize existing calls to a gather load function.
15100 (vect_analyze_data_refs): Consider using gather loads if
15101 supports_vec_gather_load_p.
15102 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
15103 (vect_get_gather_scatter_offset_type): Likewise.
15104 (vect_convert_mask_for_vectype): Likewise.
15105 (vect_add_conversion_to_patterm): Likewise.
15106 (vect_try_gather_scatter_pattern): Likewise.
15107 (vect_recog_gather_scatter_pattern): New pattern recognizer.
15108 (vect_vect_recog_func_ptrs): Add it.
15109 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15110 internal_fn_mask_index and internal_gather_scatter_fn_p.
15111 (check_load_store_masking): Take the gather_scatter_info as an
15112 argument and handle gather loads.
15113 (vect_get_gather_scatter_ops): New function.
15114 (vectorizable_call): Check internal_load_fn_p.
15115 (vectorizable_load): Likewise. Handle gather load internal
15117 (vectorizable_store): Update call to check_load_store_masking.
15118 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
15119 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
15120 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
15121 (aarch64_gather_scale_operand_d): New predicates.
15122 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
15123 (mask_gather_load<mode>): New insns.
15125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15126 Alan Hayward <alan.hayward@arm.com>
15127 David Sherwood <david.sherwood@arm.com>
15129 * optabs.def (fold_left_plus_optab): New optab.
15130 * doc/md.texi (fold_left_plus_@var{m}): Document.
15131 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
15132 * internal-fn.c (fold_left_direct): Define.
15133 (expand_fold_left_optab_fn): Likewise.
15134 (direct_fold_left_optab_supported_p): Likewise.
15135 * fold-const-call.c (fold_const_fold_left): New function.
15136 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
15137 * tree-parloops.c (valid_reduction_p): New function.
15138 (gather_scalar_reductions): Use it.
15139 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
15140 (vect_finish_replace_stmt): Declare.
15141 * tree-vect-loop.c (fold_left_reduction_fn): New function.
15142 (needs_fold_left_reduction_p): New function, split out from...
15143 (vect_is_simple_reduction): ...here. Accept reductions that
15144 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
15145 (vect_force_simple_reduction): Also store the reduction type in
15146 the assignment's STMT_VINFO_REDUC_TYPE.
15147 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
15148 (merge_with_identity): New function.
15149 (vect_expand_fold_left): Likewise.
15150 (vectorize_fold_left_reduction): Likewise.
15151 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
15152 scalar phi in place for it. Check for target support and reject
15153 cases that would reassociate the operation. Defer the transform
15154 phase to vectorize_fold_left_reduction.
15155 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
15156 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
15157 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
15159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15161 * tree-if-conv.c (predicate_mem_writes): Remove redundant
15162 call to ifc_temp_var.
15164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15165 Alan Hayward <alan.hayward@arm.com>
15166 David Sherwood <david.sherwood@arm.com>
15168 * target.def (legitimize_address_displacement): Take the original
15169 offset as a poly_int.
15170 * targhooks.h (default_legitimize_address_displacement): Update
15172 * targhooks.c (default_legitimize_address_displacement): Likewise.
15173 * doc/tm.texi: Regenerate.
15174 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
15175 as an argument, moving assert of ad->disp == ad->disp_term to...
15176 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
15177 Try calling targetm.legitimize_address_displacement before expanding
15178 the address rather than afterwards, and adjust for the new interface.
15179 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
15180 Match the new hook interface. Handle SVE addresses.
15181 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
15182 new hook interface.
15184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15186 * Makefile.in (OBJS): Add early-remat.o.
15187 * target.def (select_early_remat_modes): New hook.
15188 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
15189 * doc/tm.texi: Regenerate.
15190 * targhooks.h (default_select_early_remat_modes): Declare.
15191 * targhooks.c (default_select_early_remat_modes): New function.
15192 * timevar.def (TV_EARLY_REMAT): New timevar.
15193 * passes.def (pass_early_remat): New pass.
15194 * tree-pass.h (make_pass_early_remat): Declare.
15195 * early-remat.c: New file.
15196 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
15198 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
15200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15201 Alan Hayward <alan.hayward@arm.com>
15202 David Sherwood <david.sherwood@arm.com>
15204 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
15205 vfm1 with a bound_epilog parameter.
15206 (vect_do_peeling): Update calls accordingly, and move the prologue
15207 call earlier in the function. Treat the base bound_epilog as 0 for
15208 fully-masked loops and retain vf - 1 for other loops. Add 1 to
15209 this base when peeling for gaps.
15210 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
15211 with fully-masked loops.
15212 (vect_estimate_min_profitable_iters): Handle the single peeled
15213 iteration in that case.
15215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15216 Alan Hayward <alan.hayward@arm.com>
15217 David Sherwood <david.sherwood@arm.com>
15219 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
15220 single-element interleaving even if the size is not a power of 2.
15221 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
15222 accesses for single-element interleaving if the group size is
15225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15226 Alan Hayward <alan.hayward@arm.com>
15227 David Sherwood <david.sherwood@arm.com>
15229 * doc/md.texi (fold_extract_last_@var{m}): Document.
15230 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
15231 * optabs.def (fold_extract_last_optab): New optab.
15232 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
15233 * internal-fn.c (fold_extract_direct): New macro.
15234 (expand_fold_extract_optab_fn): Likewise.
15235 (direct_fold_extract_optab_supported_p): Likewise.
15236 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
15237 * tree-vect-loop.c (vect_model_reduction_cost): Handle
15238 EXTRACT_LAST_REDUCTION.
15239 (get_initial_def_for_reduction): Do not create an initial vector
15240 for EXTRACT_LAST_REDUCTION reductions.
15241 (vectorizable_reduction): Leave the scalar phi in place for
15242 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
15243 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
15244 epilogue code for EXTRACT_LAST_REDUCTION and defer the
15245 transform phase to vectorizable_condition.
15246 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
15248 (vect_finish_stmt_generation): ...here.
15249 (vect_finish_replace_stmt): New function.
15250 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
15251 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
15253 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
15255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15256 Alan Hayward <alan.hayward@arm.com>
15257 David Sherwood <david.sherwood@arm.com>
15259 * doc/md.texi (extract_last_@var{m}): Document.
15260 * optabs.def (extract_last_optab): New optab.
15261 * internal-fn.def (EXTRACT_LAST): New internal function.
15262 * internal-fn.c (cond_unary_direct): New macro.
15263 (expand_cond_unary_optab_fn): Likewise.
15264 (direct_cond_unary_optab_supported_p): Likewise.
15265 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
15266 loops using EXTRACT_LAST.
15267 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
15268 (extract_last_<mode>): ...this optab.
15269 (vec_extract<mode><Vel>): Update accordingly.
15271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15272 Alan Hayward <alan.hayward@arm.com>
15273 David Sherwood <david.sherwood@arm.com>
15275 * target.def (empty_mask_is_expensive): New hook.
15276 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
15277 * doc/tm.texi: Regenerate.
15278 * targhooks.h (default_empty_mask_is_expensive): Declare.
15279 * targhooks.c (default_empty_mask_is_expensive): New function.
15280 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
15281 if the target says that empty masks are expensive.
15282 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
15284 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
15286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15287 Alan Hayward <alan.hayward@arm.com>
15288 David Sherwood <david.sherwood@arm.com>
15290 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
15291 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
15292 (vect_use_loop_mask_for_alignment_p): New function.
15293 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
15294 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
15295 niters_skip argument. Make sure that the first niters_skip elements
15296 of the first iteration are inactive.
15297 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
15298 Update call to vect_set_loop_masks_directly.
15299 (get_misalign_in_elems): New function, split out from...
15300 (vect_gen_prolog_loop_niters): ...here.
15301 (vect_update_init_of_dr): Take a code argument that specifies whether
15302 the adjustment should be added or subtracted.
15303 (vect_update_init_of_drs): Likewise.
15304 (vect_prepare_for_masked_peels): New function.
15305 (vect_do_peeling): Skip prologue peeling if we're using a mask
15306 instead. Update call to vect_update_inits_of_drs.
15307 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15309 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
15310 alignment. Do not include the number of peeled iterations in
15311 the minimum threshold in that case.
15312 (vectorizable_induction): Adjust the start value down by
15313 LOOP_VINFO_MASK_SKIP_NITERS iterations.
15314 (vect_transform_loop): Call vect_prepare_for_masked_peels.
15315 Take the number of skipped iterations into account when calculating
15317 * tree-vect-stmts.c (vect_gen_while_not): New function.
15319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15320 Alan Hayward <alan.hayward@arm.com>
15321 David Sherwood <david.sherwood@arm.com>
15323 * doc/sourcebuild.texi (vect_fully_masked): Document.
15324 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
15325 default value to 0.
15326 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
15328 (vect_analyze_loop_2): ...here. Don't check the vectorization
15329 factor against the number of loop iterations if the loop is
15332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15333 Alan Hayward <alan.hayward@arm.com>
15334 David Sherwood <david.sherwood@arm.com>
15336 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
15337 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
15338 (dump_groups): Update accordingly.
15339 (iv_use::mem_type): New member variable.
15340 (address_p): New function.
15341 (record_use): Add a mem_type argument and initialize the new
15343 (record_group_use): Add a mem_type argument. Use address_p.
15344 Remove obsolete null checks of base_object. Update call to record_use.
15345 (find_interesting_uses_op): Update call to record_group_use.
15346 (find_interesting_uses_cond): Likewise.
15347 (find_interesting_uses_address): Likewise.
15348 (get_mem_type_for_internal_fn): New function.
15349 (find_address_like_use): Likewise.
15350 (find_interesting_uses_stmt): Try find_address_like_use before
15351 calling find_interesting_uses_op.
15352 (addr_offset_valid_p): Use the iv mem_type field as the type
15353 of the addressed memory.
15354 (add_autoinc_candidates): Likewise.
15355 (get_address_cost): Likewise.
15356 (split_small_address_groups_p): Use address_p.
15357 (split_address_groups): Likewise.
15358 (add_iv_candidate_for_use): Likewise.
15359 (autoinc_possible_for_pair): Likewise.
15360 (rewrite_groups): Likewise.
15361 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
15362 (determine_group_iv_cost): Update after split of USE_ADDRESS.
15363 (get_alias_ptr_type_for_ptr_address): New function.
15364 (rewrite_use_address): Rewrite address uses in calls that were
15365 identified by find_address_like_use.
15367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15368 Alan Hayward <alan.hayward@arm.com>
15369 David Sherwood <david.sherwood@arm.com>
15371 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
15373 * gimple-expr.h (is_gimple_addressable: Likewise.
15374 * gimple-expr.c (is_gimple_address): Likewise.
15375 * internal-fn.c (expand_call_mem_ref): New function.
15376 (expand_mask_load_optab_fn): Use it.
15377 (expand_mask_store_optab_fn): Likewise.
15379 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15380 Alan Hayward <alan.hayward@arm.com>
15381 David Sherwood <david.sherwood@arm.com>
15383 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
15384 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
15385 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
15386 (cond_umax@var{mode}): Document.
15387 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
15388 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
15389 (cond_umin_optab, cond_umax_optab): New optabs.
15390 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
15391 (COND_IOR, COND_XOR): New internal functions.
15392 * internal-fn.h (get_conditional_internal_fn): Declare.
15393 * internal-fn.c (cond_binary_direct): New macro.
15394 (expand_cond_binary_optab_fn): Likewise.
15395 (direct_cond_binary_optab_supported_p): Likewise.
15396 (get_conditional_internal_fn): New function.
15397 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
15398 Cope with reduction statements that are vectorized as calls rather
15400 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
15401 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
15402 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
15403 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15404 (UNSPEC_COND_EOR): New unspecs.
15405 (optab): Add mappings for them.
15406 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
15407 (sve_int_op, sve_fp_op): New int attributes.
15409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15410 Alan Hayward <alan.hayward@arm.com>
15411 David Sherwood <david.sherwood@arm.com>
15413 * optabs.def (while_ult_optab): New optab.
15414 * doc/md.texi (while_ult@var{m}@var{n}): Document.
15415 * internal-fn.def (WHILE_ULT): New internal function.
15416 * internal-fn.h (direct_internal_fn_supported_p): New override
15417 that takes two types as argument.
15418 * internal-fn.c (while_direct): New macro.
15419 (expand_while_optab_fn): New function.
15420 (convert_optab_supported_p): Likewise.
15421 (direct_while_optab_supported_p): New macro.
15422 * wide-int.h (wi::udiv_ceil): New function.
15423 * tree-vectorizer.h (rgroup_masks): New structure.
15424 (vec_loop_masks): New typedef.
15425 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
15426 and fully_masked_p.
15427 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
15428 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
15429 (vect_max_vf): New function.
15430 (slpeel_make_loop_iterate_ntimes): Delete.
15431 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
15432 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
15433 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
15434 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
15435 internal-fn.h, stor-layout.h and optabs-query.h.
15436 (vect_set_loop_mask): New function.
15437 (add_preheader_seq): Likewise.
15438 (add_header_seq): Likewise.
15439 (interleave_supported_p): Likewise.
15440 (vect_maybe_permute_loop_masks): Likewise.
15441 (vect_set_loop_masks_directly): Likewise.
15442 (vect_set_loop_condition_masked): Likewise.
15443 (vect_set_loop_condition_unmasked): New function, split out from
15444 slpeel_make_loop_iterate_ntimes.
15445 (slpeel_make_loop_iterate_ntimes): Rename to..
15446 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
15447 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
15448 (vect_do_peeling): Update call accordingly.
15449 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
15451 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15452 mask_compare_type, can_fully_mask_p and fully_masked_p.
15453 (release_vec_loop_masks): New function.
15454 (_loop_vec_info): Use it to free the loop masks.
15455 (can_produce_all_loop_masks_p): New function.
15456 (vect_get_max_nscalars_per_iter): Likewise.
15457 (vect_verify_full_masking): Likewise.
15458 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
15459 retries, and free the mask rgroups before retrying. Check loop-wide
15460 reasons for disallowing fully-masked loops. Make the final decision
15461 about whether use a fully-masked loop or not.
15462 (vect_estimate_min_profitable_iters): Do not assume that peeling
15463 for the number of iterations will be needed for fully-masked loops.
15464 (vectorizable_reduction): Disable fully-masked loops.
15465 (vectorizable_live_operation): Likewise.
15466 (vect_halve_mask_nunits): New function.
15467 (vect_double_mask_nunits): Likewise.
15468 (vect_record_loop_mask): Likewise.
15469 (vect_get_loop_mask): Likewise.
15470 (vect_transform_loop): Handle the case in which the final loop
15471 iteration might handle a partial vector. Call vect_set_loop_condition
15472 instead of slpeel_make_loop_iterate_ntimes.
15473 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
15474 (check_load_store_masking): New function.
15475 (prepare_load_store_mask): Likewise.
15476 (vectorizable_store): Handle fully-masked loops.
15477 (vectorizable_load): Likewise.
15478 (supportable_widening_operation): Use vect_halve_mask_nunits for
15480 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
15481 (vect_gen_while): New function.
15482 * config/aarch64/aarch64.md (umax<mode>3): New expander.
15483 (aarch64_uqdec<mode>): New insn.
15485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15486 Alan Hayward <alan.hayward@arm.com>
15487 David Sherwood <david.sherwood@arm.com>
15489 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
15490 (reduc_xor_scal_optab): New optabs.
15491 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
15492 (reduc_xor_scal_@var{m}): Document.
15493 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
15494 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
15495 internal functions.
15496 * fold-const-call.c (fold_const_call): Handle them.
15497 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
15498 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
15499 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
15500 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
15501 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
15502 (UNSPEC_XORV): New unspecs.
15503 (optab): Add entries for them.
15504 (BITWISEV): New int iterator.
15505 (bit_reduc_op): New int attributes.
15507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15508 Alan Hayward <alan.hayward@arm.com>
15509 David Sherwood <david.sherwood@arm.com>
15511 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
15512 * internal-fn.def (VEC_SHL_INSERT): New internal function.
15513 * optabs.def (vec_shl_insert_optab): New optab.
15514 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
15515 (duplicate_and_interleave): Likewise.
15516 * tree-vect-loop.c: Include internal-fn.h.
15517 (neutral_op_for_slp_reduction): New function, split out from
15518 get_initial_defs_for_reduction.
15519 (get_initial_def_for_reduction): Handle option 2 for variable-length
15520 vectors by loading the neutral value into a vector and then shifting
15521 the initial value into element 0.
15522 (get_initial_defs_for_reduction): Replace the code argument with
15523 the neutral value calculated by neutral_op_for_slp_reduction.
15524 Use gimple_build_vector for constant-length vectors.
15525 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
15526 but the first group_size elements have a neutral value.
15527 Use duplicate_and_interleave otherwise.
15528 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
15529 Update call to get_initial_defs_for_reduction. Handle SLP
15530 reductions for variable-length vectors by creating one vector
15531 result for each scalar result, with the elements associated
15532 with other scalar results stubbed out with the neutral value.
15533 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
15534 Require IFN_VEC_SHL_INSERT for double reductions on
15535 variable-length vectors, or SLP reductions that have
15536 a neutral value. Require can_duplicate_and_interleave_p
15537 support for variable-length unchained SLP reductions if there
15538 is no neutral value, such as for MIN/MAX reductions. Also require
15539 the number of vector elements to be a multiple of the number of
15540 SLP statements when doing variable-length unchained SLP reductions.
15541 Update call to vect_create_epilog_for_reduction.
15542 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
15543 and remove initial values.
15544 (duplicate_and_interleave): Make public.
15545 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
15546 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
15548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15549 Alan Hayward <alan.hayward@arm.com>
15550 David Sherwood <david.sherwood@arm.com>
15552 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
15553 (can_duplicate_and_interleave_p): New function.
15554 (vect_get_and_check_slp_defs): Take the vector of statements
15555 rather than just the current one. Remove excess parentheses.
15556 Restriction rejectinon of vect_constant_def and vect_external_def
15557 for variable-length vectors to boolean types, or types for which
15558 can_duplicate_and_interleave_p is false.
15559 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
15560 (duplicate_and_interleave): New function.
15561 (vect_get_constant_vectors): Use gimple_build_vector for
15562 constant-length vectors and suitable variable-length constant
15563 vectors. Use duplicate_and_interleave for other variable-length
15564 vectors. Don't defer the update when inserting new statements.
15566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15567 Alan Hayward <alan.hayward@arm.com>
15568 David Sherwood <david.sherwood@arm.com>
15570 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
15571 min_profitable_iters doesn't go negative.
15573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15574 Alan Hayward <alan.hayward@arm.com>
15575 David Sherwood <david.sherwood@arm.com>
15577 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
15578 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
15579 * optabs.def (vec_mask_load_lanes_optab): New optab.
15580 (vec_mask_store_lanes_optab): Likewise.
15581 * internal-fn.def (MASK_LOAD_LANES): New internal function.
15582 (MASK_STORE_LANES): Likewise.
15583 * internal-fn.c (mask_load_lanes_direct): New macro.
15584 (mask_store_lanes_direct): Likewise.
15585 (expand_mask_load_optab_fn): Handle masked operations.
15586 (expand_mask_load_lanes_optab_fn): New macro.
15587 (expand_mask_store_optab_fn): Handle masked operations.
15588 (expand_mask_store_lanes_optab_fn): New macro.
15589 (direct_mask_load_lanes_optab_supported_p): Likewise.
15590 (direct_mask_store_lanes_optab_supported_p): Likewise.
15591 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
15593 (vect_load_lanes_supported): Likewise.
15594 * tree-vect-data-refs.c (strip_conversion): New function.
15595 (can_group_stmts_p): Likewise.
15596 (vect_analyze_data_ref_accesses): Use it instead of checking
15597 for a pair of assignments.
15598 (vect_store_lanes_supported): Take a masked_p parameter.
15599 (vect_load_lanes_supported): Likewise.
15600 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
15601 vect_store_lanes_supported and vect_load_lanes_supported.
15602 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
15603 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
15604 parameter. Don't allow gaps for masked accesses.
15605 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
15606 and vect_load_lanes_supported.
15607 (get_load_store_type): Take a masked_p parameter and update
15608 call to get_group_load_store_type.
15609 (vectorizable_store): Update call to get_load_store_type.
15610 Handle IFN_MASK_STORE_LANES.
15611 (vectorizable_load): Update call to get_load_store_type.
15612 Handle IFN_MASK_LOAD_LANES.
15614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15615 Alan Hayward <alan.hayward@arm.com>
15616 David Sherwood <david.sherwood@arm.com>
15618 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
15620 * config/aarch64/aarch64-protos.h
15621 (aarch64_sve_struct_memory_operand_p): Declare.
15622 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
15623 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
15624 (VPRED, vpred): Handle SVE structure modes.
15625 * config/aarch64/constraints.md (Utx): New constraint.
15626 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
15627 (aarch64_sve_struct_nonimmediate_operand): New predicates.
15628 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
15629 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
15630 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
15631 structure modes. Split into pieces after RA.
15632 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
15633 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
15635 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
15636 SVE structure modes.
15637 (aarch64_classify_address): Likewise.
15638 (sizetochar): Move earlier in file.
15639 (aarch64_print_operand): Handle SVE register lists.
15640 (aarch64_array_mode): New function.
15641 (aarch64_sve_struct_memory_operand_p): Likewise.
15642 (TARGET_ARRAY_MODE): Redefine.
15644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15645 Alan Hayward <alan.hayward@arm.com>
15646 David Sherwood <david.sherwood@arm.com>
15648 * target.def (array_mode): New target hook.
15649 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
15650 * doc/tm.texi: Regenerate.
15651 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
15652 * hooks.c (hook_optmode_mode_uhwi_none): New function.
15653 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
15654 targetm.array_mode.
15655 * stor-layout.c (mode_for_array): Likewise. Support polynomial
15658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15659 Alan Hayward <alan.hayward@arm.com>
15660 David Sherwood <david.sherwood@arm.com>
15662 * fold-const.c (fold_binary_loc): Check the argument types
15663 rather than the result type when testing for a vector operation.
15665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15667 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
15668 * doc/tm.texi: Regenerate.
15670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15671 Alan Hayward <alan.hayward@arm.com>
15672 David Sherwood <david.sherwood@arm.com>
15674 * doc/invoke.texi (-msve-vector-bits=): Document new option.
15675 (sve): Document new AArch64 extension.
15676 * doc/md.texi (w): Extend the description of the AArch64
15677 constraint to include SVE vectors.
15678 (Upl, Upa): Document new AArch64 predicate constraints.
15679 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
15681 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
15682 (msve-vector-bits=): New option.
15683 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
15684 SVE when these are disabled.
15685 (sve): New extension.
15686 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
15687 modes. Adjust their number of units based on aarch64_sve_vg.
15688 (MAX_BITSIZE_MODE_ANY_MODE): Define.
15689 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
15690 aarch64_addr_query_type.
15691 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
15692 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
15693 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
15694 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
15695 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
15696 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
15697 (aarch64_simd_imm_zero_p): Delete.
15698 (aarch64_check_zero_based_sve_index_immediate): Declare.
15699 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15700 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15701 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15702 (aarch64_sve_float_mul_immediate_p): Likewise.
15703 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15704 rather than an rtx.
15705 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
15706 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
15707 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
15708 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
15709 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
15710 (aarch64_regmode_natural_size): Likewise.
15711 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
15712 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
15714 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
15715 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
15716 for VG and the SVE predicate registers.
15717 (V_ALIASES): Add a "z"-prefixed alias.
15718 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
15719 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
15720 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
15721 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
15722 (REG_CLASS_NAMES): Add entries for them.
15723 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
15724 and the predicate registers.
15725 (aarch64_sve_vg): Declare.
15726 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
15727 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
15728 (REGMODE_NATURAL_SIZE): Define.
15729 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
15731 * config/aarch64/aarch64.c: Include cfgrtl.h.
15732 (simd_immediate_info): Add a constructor for series vectors,
15733 and an associated step field.
15734 (aarch64_sve_vg): New variable.
15735 (aarch64_dbx_register_number): Handle VG and the predicate registers.
15736 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
15737 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
15738 (VEC_ANY_DATA, VEC_STRUCT): New constants.
15739 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
15740 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
15741 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
15742 (aarch64_get_mask_mode): New functions.
15743 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
15744 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
15745 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
15746 predicate modes and predicate registers. Explicitly restrict
15747 GPRs to modes of 16 bytes or smaller. Only allow FP registers
15748 to store a vector mode if it is recognized by
15749 aarch64_classify_vector_mode.
15750 (aarch64_regmode_natural_size): New function.
15751 (aarch64_hard_regno_caller_save_mode): Return the original mode
15753 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
15754 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
15755 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
15756 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
15758 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
15759 does not overlap dest if the function is frame-related. Handle
15761 (aarch64_split_add_offset): New function.
15762 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
15763 them aarch64_add_offset.
15764 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
15765 and update call to aarch64_sub_sp.
15766 (aarch64_add_cfa_expression): New function.
15767 (aarch64_expand_prologue): Pass extra temporary registers to the
15768 functions above. Handle the case in which we need to emit new
15769 DW_CFA_expressions for registers that were originally saved
15770 relative to the stack pointer, but now have to be expressed
15771 relative to the frame pointer.
15772 (aarch64_output_mi_thunk): Pass extra temporary registers to the
15774 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
15775 IP0 and IP1 values for SVE frames.
15776 (aarch64_expand_vec_series): New function.
15777 (aarch64_expand_sve_widened_duplicate): Likewise.
15778 (aarch64_expand_sve_const_vector): Likewise.
15779 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
15780 Handle SVE constants. Use emit_move_insn to move a force_const_mem
15781 into the register, rather than emitting a SET directly.
15782 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
15783 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
15784 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
15785 (offset_9bit_signed_scaled_p): New functions.
15786 (aarch64_replicate_bitmask_imm): New function.
15787 (aarch64_bitmask_imm): Use it.
15788 (aarch64_cannot_force_const_mem): Reject expressions involving
15789 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
15790 (aarch64_classify_index): Handle SVE indices, by requiring
15791 a plain register index with a scale that matches the element size.
15792 (aarch64_classify_address): Handle SVE addresses. Assert that
15793 the mode of the address is VOIDmode or an integer mode.
15794 Update call to aarch64_classify_symbol.
15795 (aarch64_classify_symbolic_expression): Update call to
15796 aarch64_classify_symbol.
15797 (aarch64_const_vec_all_in_range_p): New function.
15798 (aarch64_print_vector_float_operand): Likewise.
15799 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
15800 "vN" for FP registers with SVE modes. Handle (const ...) vectors
15801 and the FP immediates 1.0 and 0.5.
15802 (aarch64_print_address_internal): Handle SVE addresses.
15803 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
15804 (aarch64_regno_regclass): Handle predicate registers.
15805 (aarch64_secondary_reload): Handle big-endian reloads of SVE
15807 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
15808 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
15809 (aarch64_convert_sve_vector_bits): New function.
15810 (aarch64_override_options): Use it to handle -msve-vector-bits=.
15811 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
15812 rather than an rtx.
15813 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
15814 Handle SVE vector and predicate modes. Accept VL-based constants
15815 that need only one temporary register, and VL offsets that require
15816 no temporary registers.
15817 (aarch64_conditional_register_usage): Mark the predicate registers
15818 as fixed if SVE isn't available.
15819 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
15820 Return true for SVE vector and predicate modes.
15821 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
15822 rather than an unsigned int. Handle SVE modes.
15823 (aarch64_preferred_simd_mode): Update call accordingly. Handle
15825 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
15827 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
15828 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
15829 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
15830 (aarch64_sve_float_mul_immediate_p): New functions.
15831 (aarch64_sve_valid_immediate): New function.
15832 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
15833 Explicitly reject structure modes. Check for INDEX constants.
15834 Handle PTRUE and PFALSE constants.
15835 (aarch64_check_zero_based_sve_index_immediate): New function.
15836 (aarch64_simd_imm_zero_p): Delete.
15837 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
15838 vector modes. Accept constants in the range of CNT[BHWD].
15839 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
15840 ask for an Advanced SIMD mode.
15841 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
15842 (aarch64_simd_vector_alignment): Handle SVE predicates.
15843 (aarch64_vectorize_preferred_vector_alignment): New function.
15844 (aarch64_simd_vector_alignment_reachable): Use it instead of
15846 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
15847 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
15849 (MAX_VECT_LEN): Delete.
15850 (expand_vec_perm_d): Add a vec_flags field.
15851 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
15852 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
15853 (aarch64_evpc_ext): Don't apply a big-endian lane correction
15855 (aarch64_evpc_rev): Rename to...
15856 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
15857 (aarch64_evpc_rev_global): New function.
15858 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
15859 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
15861 (aarch64_evpc_sve_tbl): New function.
15862 (aarch64_expand_vec_perm_const_1): Update after rename of
15863 aarch64_evpc_rev. Handle SVE permutes too, trying
15864 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
15865 than aarch64_evpc_tbl.
15866 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
15867 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
15868 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
15869 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
15870 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
15871 (aarch64_expand_sve_vcond): New functions.
15872 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
15873 of aarch64_vector_mode_p.
15874 (aarch64_dwarf_poly_indeterminate_value): New function.
15875 (aarch64_compute_pressure_classes): Likewise.
15876 (aarch64_can_change_mode_class): Likewise.
15877 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
15878 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
15879 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
15880 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
15881 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
15882 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
15883 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
15884 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
15886 (Dn, Dl, Dr): Accept const as well as const_vector.
15887 (Dz): Likewise. Compare against CONST0_RTX.
15888 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
15889 of "vector" where appropriate.
15890 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
15891 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
15892 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
15893 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
15894 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
15895 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
15896 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
15897 (v_int_equiv): Extend to SVE modes.
15898 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
15900 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
15901 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
15902 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
15903 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
15904 (SVE_COND_FP_CMP): New int iterators.
15905 (perm_hilo): Handle the new unpack unspecs.
15906 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
15908 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
15909 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
15910 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
15911 (aarch64_equality_operator, aarch64_constant_vector_operand)
15912 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
15913 (aarch64_sve_nonimmediate_operand): Likewise.
15914 (aarch64_sve_general_operand): Likewise.
15915 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
15916 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
15917 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
15918 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
15919 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
15920 (aarch64_sve_float_arith_immediate): Likewise.
15921 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
15922 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
15923 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
15924 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
15925 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
15926 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
15927 (aarch64_sve_float_arith_operand): Likewise.
15928 (aarch64_sve_float_arith_with_sub_operand): Likewise.
15929 (aarch64_sve_float_mul_operand): Likewise.
15930 (aarch64_sve_vec_perm_operand): Likewise.
15931 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
15932 (aarch64_mov_operand): Accept const_poly_int and const_vector.
15933 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
15934 as well as const_vector.
15935 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
15936 in file. Use CONST0_RTX and CONSTM1_RTX.
15937 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
15938 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
15939 Use aarch64_simd_imm_zero.
15940 * config/aarch64/aarch64-sve.md: New file.
15941 * config/aarch64/aarch64.md: Include it.
15942 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
15943 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
15944 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
15945 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
15946 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
15947 (sve): New attribute.
15948 (enabled): Disable instructions with the sve attribute unless
15950 (movqi, movhi): Pass CONST_POLY_INT operaneds through
15951 aarch64_expand_mov_immediate.
15952 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
15953 CNT[BHSD] immediates.
15954 (movti): Split CONST_POLY_INT moves into two halves.
15955 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
15956 Split additions that need a temporary here if the destination
15957 is the stack pointer.
15958 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
15959 (*add<mode>3_poly_1): New instruction.
15960 (set_clobber_cc): New expander.
15962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15964 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
15965 parameter and use it instead of GET_MODE_SIZE (innermode). Use
15966 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
15967 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
15968 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
15969 Change innermode from fixed_mode_size to machine_mode.
15970 (simplify_subreg): Update call accordingly. Handle a constant-sized
15971 subreg of a variable-length CONST_VECTOR.
15973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15974 Alan Hayward <alan.hayward@arm.com>
15975 David Sherwood <david.sherwood@arm.com>
15977 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
15978 (add_offset_to_base): New function, split out from...
15979 (create_mem_ref): ...here. When handling a scale other than 1,
15980 check first whether the address is valid without the offset.
15981 Add it into the base if so, leaving the index and scale as-is.
15983 2018-01-12 Jakub Jelinek <jakub@redhat.com>
15986 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
15987 fold_for_warn before checking if arg2 is INTEGER_CST.
15989 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
15991 * config/rs6000/predicates.md (load_multiple_operation): Delete.
15992 (store_multiple_operation): Delete.
15993 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
15994 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
15995 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
15996 guarded by TARGET_STRING.
15997 (rs6000_output_load_multiple): Delete.
15998 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
15999 OPTION_MASK_STRING / TARGET_STRING handling.
16000 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16001 (const rs6000_opt_masks) <"string">: Change mask to 0.
16002 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16003 (MASK_STRING): Delete.
16004 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16006 (load_multiple): Delete.
16013 (store_multiple): Delete.
16020 (movmemsi_8reg): Delete.
16021 (corresponding unnamed define_insn): Delete.
16022 (movmemsi_6reg): Delete.
16023 (corresponding unnamed define_insn): Delete.
16024 (movmemsi_4reg): Delete.
16025 (corresponding unnamed define_insn): Delete.
16026 (movmemsi_2reg): Delete.
16027 (corresponding unnamed define_insn): Delete.
16028 (movmemsi_1reg): Delete.
16029 (corresponding unnamed define_insn): Delete.
16030 * config/rs6000/rs6000.opt (mno-string): New.
16031 (mstring): Replace by deprecation warning stub.
16032 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16034 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16036 * regrename.c (regrename_do_replace): If replacing the same
16037 reg multiple times, try to reuse last created gen_raw_REG.
16040 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16041 main to workaround a bug in GDB.
16043 2018-01-12 Tom de Vries <tom@codesourcery.com>
16046 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16048 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16050 PR rtl-optimization/80481
16051 * ira-color.c (get_cap_member): New function.
16052 (allocnos_conflict_by_live_ranges_p): Use it.
16053 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16054 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16056 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16059 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16060 (*saddl_se_1): Ditto.
16061 (*ssubsi_1): Ditto.
16062 (*ssubl_se_1): Ditto.
16064 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16066 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
16067 rather than wi::to_widest for DR_INITs.
16068 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
16069 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
16070 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
16072 (vect_analyze_group_access_1): Note that here.
16074 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16076 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
16077 polynomial type sizes.
16079 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
16081 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
16082 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
16083 (gimple_add_tmp_var): Likewise.
16085 2018-01-12 Martin Liska <mliska@suse.cz>
16087 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
16088 (gimple_alloc_sizes): Likewise.
16089 (dump_gimple_statistics): Use PRIu64 in printf format.
16090 * gimple.h: Change uint64_t to int.
16092 2018-01-12 Martin Liska <mliska@suse.cz>
16094 * tree-core.h: Use uint64_t instead of int.
16095 * tree.c (tree_node_counts): Likewise.
16096 (tree_node_sizes): Likewise.
16097 (dump_tree_statistics): Use PRIu64 in printf format.
16099 2018-01-12 Martin Liska <mliska@suse.cz>
16101 * Makefile.in: As qsort_chk is implemented in vec.c, add
16102 vec.o to linkage of gencfn-macros.
16103 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
16104 passing the info to record_node_allocation_statistics.
16105 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
16107 * ggc-common.c (struct ggc_usage): Add operator== and use
16108 it in operator< and compare function.
16109 * mem-stats.h (struct mem_usage): Likewise.
16110 * vec.c (struct vec_usage): Remove operator< and compare
16111 function. Can be simply inherited.
16113 2018-01-12 Martin Jambor <mjambor@suse.cz>
16116 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
16117 * tree-ssa-math-opts.c: Include domwalk.h.
16118 (convert_mult_to_fma_1): New function.
16119 (fma_transformation_info): New type.
16120 (fma_deferring_state): Likewise.
16121 (cancel_fma_deferring): New function.
16122 (result_of_phi): Likewise.
16123 (last_fma_candidate_feeds_initial_phi): Likewise.
16124 (convert_mult_to_fma): Added deferring logic, split actual
16125 transformation to convert_mult_to_fma_1.
16126 (math_opts_dom_walker): New type.
16127 (math_opts_dom_walker::after_dom_children): New method, body moved
16128 here from pass_optimize_widening_mul::execute, added deferring logic
16130 (pass_optimize_widening_mul::execute): Moved most of code to
16131 math_opts_dom_walker::after_dom_children.
16132 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
16133 * config/i386/i386.c (ix86_option_override_internal): Added
16134 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
16136 2018-01-12 Richard Biener <rguenther@suse.de>
16139 * dwarf2out.c (gen_variable_die): Do not reset old_die for
16140 inline instance vars.
16142 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
16145 * config/rx/rx.c (rx_is_restricted_memory_address):
16146 Handle SUBREG case.
16148 2018-01-12 Richard Biener <rguenther@suse.de>
16150 PR tree-optimization/80846
16151 * target.def (split_reduction): New target hook.
16152 * targhooks.c (default_split_reduction): New function.
16153 * targhooks.h (default_split_reduction): Declare.
16154 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
16155 target requests first reduce vectors by combining low and high
16157 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
16158 (get_vectype_for_scalar_type_and_size): Export.
16159 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
16160 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
16161 * doc/tm.texi: Regenerate.
16162 * config/i386/i386.c (ix86_split_reduction): Implement
16163 TARGET_VECTORIZE_SPLIT_REDUCTION.
16165 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16168 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
16169 in PIC mode except for TARGET_VXWORKS_RTP.
16170 * config/sparc/sparc.c: Include cfgrtl.h.
16171 (TARGET_INIT_PIC_REG): Define.
16172 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
16173 (sparc_pic_register_p): New predicate.
16174 (sparc_legitimate_address_p): Use it.
16175 (sparc_legitimize_pic_address): Likewise.
16176 (sparc_delegitimize_address): Likewise.
16177 (sparc_mode_dependent_address_p): Likewise.
16178 (gen_load_pcrel_sym): Remove 4th parameter.
16179 (load_got_register): Adjust call to above. Remove obsolete stuff.
16180 (sparc_expand_prologue): Do not call load_got_register here.
16181 (sparc_flat_expand_prologue): Likewise.
16182 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
16183 (sparc_use_pseudo_pic_reg): New function.
16184 (sparc_init_pic_reg): Likewise.
16185 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16186 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
16188 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
16190 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
16191 Add item for branch_cost.
16193 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16195 PR rtl-optimization/83565
16196 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
16197 not extend the result to a larger mode for rotate operations.
16198 (num_sign_bit_copies1): Likewise.
16200 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16203 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
16205 Use values-Xc.o for -pedantic.
16206 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
16208 2018-01-12 Martin Liska <mliska@suse.cz>
16211 * ipa-devirt.c (final_warning_record::grow_type_warnings):
16213 (possible_polymorphic_call_targets): Use it.
16214 (ipa_devirt): Likewise.
16216 2018-01-12 Martin Liska <mliska@suse.cz>
16218 * profile-count.h (enum profile_quality): Use 0 as invalid
16219 enum value of profile_quality.
16221 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
16223 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
16224 -mext-string options.
16226 2018-01-12 Richard Biener <rguenther@suse.de>
16228 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
16229 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
16230 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
16232 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
16234 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
16236 * configure.ac (--with-long-double-format): Add support for the
16237 configuration option to change the default long double format on
16239 * config.gcc (powerpc*-linux*-*): Likewise.
16240 * configure: Regenerate.
16241 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
16242 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
16243 used without modification.
16245 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16247 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
16248 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
16249 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
16250 MISC_BUILTIN_SPEC_BARRIER.
16251 (rs6000_init_builtins): Likewise.
16252 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
16254 (speculation_barrier): New define_insn.
16255 * doc/extend.texi: Document __builtin_speculation_barrier.
16257 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16260 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
16261 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
16262 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
16264 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
16265 integral modes instead of "ss" and "sd".
16266 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
16267 vectors with 32-bit and 64-bit elements.
16268 (vecdupssescalarmodesuffix): New mode attribute.
16269 (vec_dup<mode>): Use it.
16271 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
16274 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
16275 frame if argument is passed on stack.
16277 2018-01-11 Jakub Jelinek <jakub@redhat.com>
16280 * ree.c (combine_reaching_defs): Optimize also
16281 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
16282 reg2=any_extend(exp); reg1=reg2;, formatting fix.
16284 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16286 PR middle-end/83189
16287 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
16289 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
16291 PR middle-end/83718
16292 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
16293 after they are computed.
16295 2018-01-11 Bin Cheng <bin.cheng@arm.com>
16297 PR tree-optimization/83695
16298 * gimple-loop-linterchange.cc
16299 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
16300 reset cached scev information after interchange.
16301 (pass_linterchange::execute): Remove call to scev_reset_htab.
16303 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16305 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
16306 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
16307 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
16308 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
16309 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
16310 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
16311 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
16312 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
16313 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
16314 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
16315 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
16316 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
16317 (V_lane_reg): Likewise.
16318 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
16320 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
16321 (vfmal_lane_low<mode>_intrinsic,
16322 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
16323 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
16324 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
16325 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
16326 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
16327 vfmsl_lane_high<mode>_intrinsic): New define_insns.
16329 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16331 * config/arm/arm-cpus.in (fp16fml): New feature.
16332 (ALL_SIMD): Add fp16fml.
16333 (armv8.2-a): Add fp16fml as an option.
16334 (armv8.3-a): Likewise.
16335 (armv8.4-a): Add fp16fml as part of fp16.
16336 * config/arm/arm.h (TARGET_FP16FML): Define.
16337 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
16339 * config/arm/arm-modes.def (V2HF): Define.
16340 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
16341 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
16342 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
16343 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
16344 vfmsl_low, vfmsl_high): New set of builtins.
16345 * config/arm/iterators.md (PLUSMINUS): New code iterator.
16346 (vfml_op): New code attribute.
16347 (VFMLHALVES): New int iterator.
16348 (VFML, VFMLSEL): New mode attributes.
16349 (V_reg): Define mapping for V2HF.
16350 (V_hi, V_lo): New mode attributes.
16351 (VF_constraint): Likewise.
16352 (vfml_half, vfml_half_selector): New int attributes.
16353 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
16355 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
16356 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
16358 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
16359 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
16360 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
16361 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
16363 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
16364 Document new effective target and option set.
16366 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16368 * config/arm/arm-cpus.in (armv8_4): New feature.
16369 (ARMv8_4a): New fgroup.
16370 (armv8.4-a): New arch.
16371 * config/arm/arm-tables.opt: Regenerate.
16372 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
16373 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
16374 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
16375 Add matching rules for -march=armv8.4-a and extensions.
16376 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
16378 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
16381 * config/rx/rx.md (BW): New mode attribute.
16382 (sync_lock_test_and_setsi): Add mode suffix to insn output.
16384 2018-01-11 Richard Biener <rguenther@suse.de>
16386 PR tree-optimization/83435
16387 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
16388 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
16389 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
16391 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16392 Alan Hayward <alan.hayward@arm.com>
16393 David Sherwood <david.sherwood@arm.com>
16395 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
16397 (aarch64_classify_address): Initialize it. Track polynomial offsets.
16398 (aarch64_print_address_internal): Use it to check for a zero offset.
16400 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16401 Alan Hayward <alan.hayward@arm.com>
16402 David Sherwood <david.sherwood@arm.com>
16404 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
16405 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
16406 Return a poly_int64 rather than a HOST_WIDE_INT.
16407 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
16408 rather than a HOST_WIDE_INT.
16409 * config/aarch64/aarch64.h (aarch64_frame): Protect with
16410 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
16411 hard_fp_offset, frame_size, initial_adjust, callee_offset and
16412 final_offset from HOST_WIDE_INT to poly_int64.
16413 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
16414 to_constant when getting the number of units in an Advanced SIMD
16416 (aarch64_builtin_vectorized_function): Check for a constant number
16418 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
16420 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
16421 attribute instead of GET_MODE_NUNITS.
16422 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
16423 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
16424 GET_MODE_SIZE for fixed-size registers.
16425 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
16426 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
16427 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
16428 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
16429 (aarch64_print_operand, aarch64_print_address_internal)
16430 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
16431 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
16432 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
16433 Handle polynomial GET_MODE_SIZE.
16434 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
16435 wider than SImode without modification.
16436 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
16437 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
16438 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
16439 passing and returning SVE modes.
16440 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
16441 rather than GEN_INT.
16442 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
16443 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
16444 (aarch64_allocate_and_probe_stack_space): Likewise.
16445 (aarch64_layout_frame): Cope with polynomial offsets.
16446 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
16447 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
16448 polynomial offsets.
16449 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
16450 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
16451 poly_int64 rather than a HOST_WIDE_INT.
16452 (aarch64_get_separate_components, aarch64_process_components)
16453 (aarch64_expand_prologue, aarch64_expand_epilogue)
16454 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
16455 (aarch64_anchor_offset): New function, split out from...
16456 (aarch64_legitimize_address): ...here.
16457 (aarch64_builtin_vectorization_cost): Handle polynomial
16458 TYPE_VECTOR_SUBPARTS.
16459 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
16461 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
16462 number of elements from the PARALLEL rather than the mode.
16463 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
16464 rather than GET_MODE_BITSIZE.
16465 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
16466 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
16467 (aarch64_expand_vec_perm_const_1): Handle polynomial
16468 d->perm.length () and d->perm elements.
16469 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
16470 Apply to_constant to d->perm elements.
16471 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
16472 polynomial CONST_VECTOR_NUNITS.
16473 (aarch64_move_pointer): Take amount as a poly_int64 rather
16475 (aarch64_progress_pointer): Avoid temporary variable.
16476 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
16477 the mode attribute instead of GET_MODE.
16479 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16480 Alan Hayward <alan.hayward@arm.com>
16481 David Sherwood <david.sherwood@arm.com>
16483 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
16484 x exists before using it.
16485 (aarch64_add_constant_internal): Rename to...
16486 (aarch64_add_offset_1): ...this. Replace regnum with separate
16487 src and dest rtxes. Handle the case in which they're different,
16488 including when the offset is zero. Replace scratchreg with an rtx.
16489 Use 2 additions if there is no spare register into which we can
16490 move a 16-bit constant.
16491 (aarch64_add_constant): Delete.
16492 (aarch64_add_offset): Replace reg with separate src and dest
16493 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
16494 Use aarch64_add_offset_1.
16495 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
16496 an rtx rather than an int. Take the delta as a poly_int64
16497 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
16498 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
16499 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
16500 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
16501 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
16502 and aarch64_add_sp.
16503 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
16504 aarch64_add_constant.
16506 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16508 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
16509 Use scalar_float_mode.
16511 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16513 * config/aarch64/aarch64-simd.md
16514 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
16515 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
16516 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
16517 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
16518 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
16519 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
16520 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
16521 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
16522 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
16523 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
16525 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16528 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
16529 targ_options->x_arm_arch_string is non NULL.
16531 2018-01-11 Tamar Christina <tamar.christina@arm.com>
16533 * config/aarch64/aarch64.h
16534 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
16536 2018-01-11 Sudakshina Das <sudi.das@arm.com>
16539 * expmed.c (emit_store_flag_force): Swap if const op0
16540 and change VOIDmode to mode of op0.
16542 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
16544 PR rtl-optimization/83761
16545 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
16546 than bytes to mode_for_size.
16548 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16550 PR middle-end/83189
16551 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
16552 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
16555 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16557 PR middle-end/83575
16558 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
16559 when in layout mode.
16560 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
16561 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
16564 2018-01-10 Michael Collison <michael.collison@arm.com>
16566 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
16567 * config/aarch64/aarch64-option-extension.def: Add
16568 AARCH64_OPT_EXTENSION of 'fp16fml'.
16569 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16570 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
16571 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
16572 * config/aarch64/constraints.md (Ui7): New constraint.
16573 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
16574 (VFMLA_SEL_W): Ditto.
16577 (VFMLA16_LOW): New int iterator.
16578 (VFMLA16_HIGH): Ditto.
16579 (UNSPEC_FMLAL): New unspec.
16580 (UNSPEC_FMLSL): Ditto.
16581 (UNSPEC_FMLAL2): Ditto.
16582 (UNSPEC_FMLSL2): Ditto.
16583 (f16mac): New code attribute.
16584 * config/aarch64/aarch64-simd-builtins.def
16585 (aarch64_fmlal_lowv2sf): Ditto.
16586 (aarch64_fmlsl_lowv2sf): Ditto.
16587 (aarch64_fmlalq_lowv4sf): Ditto.
16588 (aarch64_fmlslq_lowv4sf): Ditto.
16589 (aarch64_fmlal_highv2sf): Ditto.
16590 (aarch64_fmlsl_highv2sf): Ditto.
16591 (aarch64_fmlalq_highv4sf): Ditto.
16592 (aarch64_fmlslq_highv4sf): Ditto.
16593 (aarch64_fmlal_lane_lowv2sf): Ditto.
16594 (aarch64_fmlsl_lane_lowv2sf): Ditto.
16595 (aarch64_fmlal_laneq_lowv2sf): Ditto.
16596 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
16597 (aarch64_fmlalq_lane_lowv4sf): Ditto.
16598 (aarch64_fmlsl_lane_lowv4sf): Ditto.
16599 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
16600 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
16601 (aarch64_fmlal_lane_highv2sf): Ditto.
16602 (aarch64_fmlsl_lane_highv2sf): Ditto.
16603 (aarch64_fmlal_laneq_highv2sf): Ditto.
16604 (aarch64_fmlsl_laneq_highv2sf): Ditto.
16605 (aarch64_fmlalq_lane_highv4sf): Ditto.
16606 (aarch64_fmlsl_lane_highv4sf): Ditto.
16607 (aarch64_fmlalq_laneq_highv4sf): Ditto.
16608 (aarch64_fmlsl_laneq_highv4sf): Ditto.
16609 * config/aarch64/aarch64-simd.md:
16610 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
16611 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16612 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
16613 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
16614 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
16615 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
16616 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
16617 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
16618 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
16619 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
16620 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
16621 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
16622 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
16623 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
16624 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
16625 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
16626 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
16627 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
16628 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
16629 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
16630 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
16631 (vfmlsl_low_u32): Ditto.
16632 (vfmlalq_low_u32): Ditto.
16633 (vfmlslq_low_u32): Ditto.
16634 (vfmlal_high_u32): Ditto.
16635 (vfmlsl_high_u32): Ditto.
16636 (vfmlalq_high_u32): Ditto.
16637 (vfmlslq_high_u32): Ditto.
16638 (vfmlal_lane_low_u32): Ditto.
16639 (vfmlsl_lane_low_u32): Ditto.
16640 (vfmlal_laneq_low_u32): Ditto.
16641 (vfmlsl_laneq_low_u32): Ditto.
16642 (vfmlalq_lane_low_u32): Ditto.
16643 (vfmlslq_lane_low_u32): Ditto.
16644 (vfmlalq_laneq_low_u32): Ditto.
16645 (vfmlslq_laneq_low_u32): Ditto.
16646 (vfmlal_lane_high_u32): Ditto.
16647 (vfmlsl_lane_high_u32): Ditto.
16648 (vfmlal_laneq_high_u32): Ditto.
16649 (vfmlsl_laneq_high_u32): Ditto.
16650 (vfmlalq_lane_high_u32): Ditto.
16651 (vfmlslq_lane_high_u32): Ditto.
16652 (vfmlalq_laneq_high_u32): Ditto.
16653 (vfmlslq_laneq_high_u32): Ditto.
16654 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
16655 (AARCH64_FL_FOR_ARCH8_4): New.
16656 (AARCH64_ISA_F16FML): New ISA flag.
16657 (TARGET_F16FML): New feature flag for fp16fml.
16658 (doc/invoke.texi): Document new fp16fml option.
16660 2018-01-10 Michael Collison <michael.collison@arm.com>
16662 * config/aarch64/aarch64-builtins.c:
16663 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
16664 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16665 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
16666 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
16667 (AARCH64_ISA_SHA3): New ISA flag.
16668 (TARGET_SHA3): New feature flag for sha3.
16669 * config/aarch64/iterators.md (sha512_op): New int attribute.
16670 (CRYPTO_SHA512): New int iterator.
16671 (UNSPEC_SHA512H): New unspec.
16672 (UNSPEC_SHA512H2): Ditto.
16673 (UNSPEC_SHA512SU0): Ditto.
16674 (UNSPEC_SHA512SU1): Ditto.
16675 * config/aarch64/aarch64-simd-builtins.def
16676 (aarch64_crypto_sha512hqv2di): New builtin.
16677 (aarch64_crypto_sha512h2qv2di): Ditto.
16678 (aarch64_crypto_sha512su0qv2di): Ditto.
16679 (aarch64_crypto_sha512su1qv2di): Ditto.
16680 (aarch64_eor3qv8hi): Ditto.
16681 (aarch64_rax1qv2di): Ditto.
16682 (aarch64_xarqv2di): Ditto.
16683 (aarch64_bcaxqv8hi): Ditto.
16684 * config/aarch64/aarch64-simd.md:
16685 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
16686 (aarch64_crypto_sha512su0qv2di): Ditto.
16687 (aarch64_crypto_sha512su1qv2di): Ditto.
16688 (aarch64_eor3qv8hi): Ditto.
16689 (aarch64_rax1qv2di): Ditto.
16690 (aarch64_xarqv2di): Ditto.
16691 (aarch64_bcaxqv8hi): Ditto.
16692 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
16693 (vsha512h2q_u64): Ditto.
16694 (vsha512su0q_u64): Ditto.
16695 (vsha512su1q_u64): Ditto.
16696 (veor3q_u16): Ditto.
16697 (vrax1q_u64): Ditto.
16698 (vxarq_u64): Ditto.
16699 (vbcaxq_u16): Ditto.
16700 * config/arm/types.md (crypto_sha512): New type attribute.
16701 (crypto_sha3): Ditto.
16702 (doc/invoke.texi): Document new sha3 option.
16704 2018-01-10 Michael Collison <michael.collison@arm.com>
16706 * config/aarch64/aarch64-builtins.c:
16707 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
16708 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16709 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
16710 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
16711 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
16712 (AARCH64_ISA_SM4): New ISA flag.
16713 (TARGET_SM4): New feature flag for sm4.
16714 * config/aarch64/aarch64-simd-builtins.def
16715 (aarch64_sm3ss1qv4si): Ditto.
16716 (aarch64_sm3tt1aq4si): Ditto.
16717 (aarch64_sm3tt1bq4si): Ditto.
16718 (aarch64_sm3tt2aq4si): Ditto.
16719 (aarch64_sm3tt2bq4si): Ditto.
16720 (aarch64_sm3partw1qv4si): Ditto.
16721 (aarch64_sm3partw2qv4si): Ditto.
16722 (aarch64_sm4eqv4si): Ditto.
16723 (aarch64_sm4ekeyqv4si): Ditto.
16724 * config/aarch64/aarch64-simd.md:
16725 (aarch64_sm3ss1qv4si): Ditto.
16726 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
16727 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
16728 (aarch64_sm4eqv4si): Ditto.
16729 (aarch64_sm4ekeyqv4si): Ditto.
16730 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
16731 (sm3part_op): Ditto.
16732 (CRYPTO_SM3TT): Ditto.
16733 (CRYPTO_SM3PART): Ditto.
16734 (UNSPEC_SM3SS1): New unspec.
16735 (UNSPEC_SM3TT1A): Ditto.
16736 (UNSPEC_SM3TT1B): Ditto.
16737 (UNSPEC_SM3TT2A): Ditto.
16738 (UNSPEC_SM3TT2B): Ditto.
16739 (UNSPEC_SM3PARTW1): Ditto.
16740 (UNSPEC_SM3PARTW2): Ditto.
16741 (UNSPEC_SM4E): Ditto.
16742 (UNSPEC_SM4EKEY): Ditto.
16743 * config/aarch64/constraints.md (Ui2): New constraint.
16744 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
16745 * config/arm/types.md (crypto_sm3): New type attribute.
16746 (crypto_sm4): Ditto.
16747 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
16748 (vsm3tt1aq_u32): Ditto.
16749 (vsm3tt1bq_u32): Ditto.
16750 (vsm3tt2aq_u32): Ditto.
16751 (vsm3tt2bq_u32): Ditto.
16752 (vsm3partw1q_u32): Ditto.
16753 (vsm3partw2q_u32): Ditto.
16754 (vsm4eq_u32): Ditto.
16755 (vsm4ekeyq_u32): Ditto.
16756 (doc/invoke.texi): Document new sm4 option.
16758 2018-01-10 Michael Collison <michael.collison@arm.com>
16760 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
16761 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
16762 (AARCH64_FL_FOR_ARCH8_4): New.
16763 (AARCH64_FL_V8_4): New flag.
16764 (doc/invoke.texi): Document new armv8.4-a option.
16766 2018-01-10 Michael Collison <michael.collison@arm.com>
16768 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
16769 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
16770 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
16771 * config/aarch64/aarch64-option-extension.def: Add
16772 AARCH64_OPT_EXTENSION of 'sha2'.
16773 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
16774 (crypto): Disable sha2 and aes if crypto disabled.
16775 (crypto): Enable aes and sha2 if enabled.
16776 (simd): Disable sha2 and aes if simd disabled.
16777 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
16779 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
16780 (TARGET_SHA2): New feature flag for sha2.
16781 (TARGET_AES): New feature flag for aes.
16782 * config/aarch64/aarch64-simd.md:
16783 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
16784 conditional on TARGET_AES.
16785 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
16786 (aarch64_crypto_sha1hsi): Make pattern conditional
16788 (aarch64_crypto_sha1hv4si): Ditto.
16789 (aarch64_be_crypto_sha1hv4si): Ditto.
16790 (aarch64_crypto_sha1su1v4si): Ditto.
16791 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
16792 (aarch64_crypto_sha1su0v4si): Ditto.
16793 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
16794 (aarch64_crypto_sha256su0v4si): Ditto.
16795 (aarch64_crypto_sha256su1v4si): Ditto.
16796 (doc/invoke.texi): Document new aes and sha2 options.
16798 2018-01-10 Martin Sebor <msebor@redhat.com>
16800 PR tree-optimization/83781
16801 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
16804 2018-01-11 Martin Sebor <msebor@gmail.com>
16805 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16807 PR tree-optimization/83501
16808 PR tree-optimization/81703
16810 * tree-ssa-strlen.c (get_string_cst): Rename...
16811 (get_string_len): ...to this. Handle global constants.
16812 (handle_char_store): Adjust.
16814 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
16815 Jim Wilson <jimw@sifive.com>
16817 * config/riscv/riscv-protos.h (riscv_output_return): New.
16818 * config/riscv/riscv.c (struct machine_function): New naked_p field.
16819 (riscv_attribute_table, riscv_output_return),
16820 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
16821 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
16822 (riscv_compute_frame_info): Only compute frame->mask if not a naked
16824 (riscv_expand_prologue): Add early return for naked function.
16825 (riscv_expand_epilogue): Likewise.
16826 (riscv_function_ok_for_sibcall): Return false for naked function.
16827 (riscv_set_current_function): New.
16828 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
16829 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
16830 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
16831 * doc/extend.texi (RISC-V Function Attributes): New.
16833 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
16835 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
16836 check for 128-bit long double before checking TCmode.
16837 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
16838 128-bit long doubles before checking TFmode or TCmode.
16839 (FLOAT128_IBM_P): Likewise.
16841 2018-01-10 Martin Sebor <msebor@redhat.com>
16843 PR tree-optimization/83671
16844 * builtins.c (c_strlen): Unconditionally return zero for the empty
16846 Use -Warray-bounds for warnings.
16847 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
16848 for non-constant array indices with COMPONENT_REF, arrays of
16849 arrays, and pointers to arrays.
16850 (gimple_fold_builtin_strlen): Determine and set length range for
16851 non-constant character arrays.
16853 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
16855 PR middle-end/81897
16856 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
16859 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
16861 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
16863 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16866 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
16867 VECTOR_MEM_ALTIVEC_OR_VSX_P.
16868 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
16869 indexed_or_indirect_operand predicate.
16870 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
16871 (*vsx_le_perm_load_v8hi): Likewise.
16872 (*vsx_le_perm_load_v16qi): Likewise.
16873 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
16874 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
16875 (*vsx_le_perm_store_v8hi): Likewise.
16876 (*vsx_le_perm_store_v16qi): Likewise.
16877 (eight unnamed splitters): Likewise.
16879 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
16881 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
16882 * config/rs6000/emmintrin.h: Likewise.
16883 * config/rs6000/mmintrin.h: Likewise.
16884 * config/rs6000/xmmintrin.h: Likewise.
16886 2018-01-10 David Malcolm <dmalcolm@redhat.com>
16889 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
16891 * tree.c (tree_nop_conversion): Return true for location wrapper
16893 (maybe_wrap_with_location): New function.
16894 (selftest::check_strip_nops): New function.
16895 (selftest::test_location_wrappers): New function.
16896 (selftest::tree_c_tests): Call it.
16897 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
16898 (maybe_wrap_with_location): New decl.
16899 (EXPR_LOCATION_WRAPPER_P): New macro.
16900 (location_wrapper_p): New inline function.
16901 (tree_strip_any_location_wrapper): New inline function.
16903 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
16906 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
16907 stack_realign_offset for the largest alignment of stack slot
16909 (ix86_find_max_used_stack_alignment): New function.
16910 (ix86_finalize_stack_frame_flags): Use it. Set
16911 max_used_stack_alignment if we don't realign stack.
16912 * config/i386/i386.h (machine_function): Add
16913 max_used_stack_alignment.
16915 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
16917 * config/arm/arm.opt (-mbranch-cost): New option.
16918 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
16921 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
16924 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
16925 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
16927 2018-01-10 Richard Biener <rguenther@suse.de>
16930 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
16931 early out so it also covers the case where we have a non-NULL
16934 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16936 PR tree-optimization/83753
16937 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
16938 for non-strided grouped accesses if the number of elements is 1.
16940 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
16943 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
16944 * i386.h (TARGET_USE_GATHER): Define.
16945 * x86-tune.def (X86_TUNE_USE_GATHER): New.
16947 2018-01-10 Martin Liska <mliska@suse.cz>
16950 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
16951 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
16953 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
16954 CLEANUP_NO_PARTITIONING is not set.
16956 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
16958 * doc/rtl.texi: Remove documentation of (const ...) wrappers
16959 for vectors, as a partial revert of r254296.
16960 * rtl.h (const_vec_p): Delete.
16961 (const_vec_duplicate_p): Don't test for vector CONSTs.
16962 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
16963 * expmed.c (make_tree): Likewise.
16966 * common.md (E, F): Use CONSTANT_P instead of checking for
16968 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
16969 checking for CONST_VECTOR.
16971 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
16973 PR middle-end/83575
16974 * predict.c (force_edge_cold): Handle in more sane way edges
16975 with no prediction.
16977 2018-01-09 Carl Love <cel@us.ibm.com>
16979 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
16981 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
16982 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
16983 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
16984 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
16985 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
16986 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
16987 * config/rs6000/rs6000-protos.h: Add extern defition for
16988 rs6000_generate_float2_double_code.
16989 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
16991 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
16992 (float2_v2df): Add define_expand.
16994 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
16997 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
16998 op_mode in the force_to_mode call.
17000 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17002 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17003 instead of checking each element individually.
17004 (aarch64_evpc_uzp): Likewise.
17005 (aarch64_evpc_zip): Likewise.
17006 (aarch64_evpc_ext): Likewise.
17007 (aarch64_evpc_rev): Likewise.
17008 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17009 instead of checking each element individually. Return true without
17011 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17012 whether all selected elements come from the same input, instead of
17013 checking each element individually. Remove calls to gen_rtx_REG,
17014 start_sequence and end_sequence and instead assert that no rtl is
17017 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17019 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17020 order of HIGH and CONST checks.
17022 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17024 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17025 if the destination isn't an SSA_NAME.
17027 2018-01-09 Richard Biener <rguenther@suse.de>
17029 PR tree-optimization/83668
17030 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17032 (canonicalize_loop_form): ... here, renamed from ...
17033 (canonicalize_loop_closed_ssa_form): ... this and amended to
17034 swap successor edges for loop exit blocks to make us use
17035 the RPO order we need for initial schedule generation.
17037 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17039 PR tree-optimization/64811
17040 * match.pd: When optimizing comparisons with Inf, avoid
17041 introducing or losing exceptions from comparisons with NaN.
17043 2018-01-09 Martin Liska <mliska@suse.cz>
17046 * asan.c (shadow_mem_size): Add gcc_assert.
17048 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17050 Don't save registers in main().
17053 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17054 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17055 * config/avr/avr.c (avr_set_current_function): Don't error if
17056 naked, OS_task or OS_main are specified at the same time.
17057 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17059 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17061 * common/config/avr/avr-common.c (avr_option_optimization_table):
17062 Switch on -mmain-is-OS_task for optimizing compilations.
17064 2018-01-09 Richard Biener <rguenther@suse.de>
17066 PR tree-optimization/83572
17067 * graphite.c: Include cfganal.h.
17068 (graphite_transform_loops): Connect infinite loops to exit
17069 and remove fake edges at the end.
17071 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17073 * ipa-inline.c (edge_badness): Revert accidental checkin.
17075 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17078 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
17079 symbols; not inline clones.
17081 2018-01-09 Jakub Jelinek <jakub@redhat.com>
17084 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
17085 hard registers. Formatting fixes.
17087 PR preprocessor/83722
17088 * gcc.c (try_generate_repro): Pass
17089 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
17090 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
17093 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
17094 Kito Cheng <kito.cheng@gmail.com>
17096 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
17097 (riscv_leaf_function_p): Delete.
17098 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
17100 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17102 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
17104 (do_ifelse): New function.
17105 (do_isel): New function.
17106 (do_sub3): New function.
17107 (do_add3): New function.
17108 (do_load_mask_compare): New function.
17109 (do_overlap_load_compare): New function.
17110 (expand_compare_loop): New function.
17111 (expand_block_compare): Call expand_compare_loop() when appropriate.
17112 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
17113 option description.
17114 (-mblock-compare-inline-loop-limit): New option.
17116 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17119 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
17120 Reverse order of second and third operands in first alternative.
17121 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
17122 of first and second elements in UNSPEC_VPERMR vector.
17123 (altivec_expand_vec_perm_le): Likewise.
17125 2018-01-08 Jeff Law <law@redhat.com>
17127 PR rtl-optimizatin/81308
17128 * tree-switch-conversion.c (cfg_altered): New file scoped static.
17129 (process_switch): If group_case_labels makes a change, then set
17131 (pass_convert_switch::execute): If a switch is converted, then
17132 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
17134 PR rtl-optimization/81308
17135 * recog.c (split_all_insns): Conditionally cleanup the CFG after
17138 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
17140 PR target/83663 - Revert r255946
17141 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
17142 generation for cases where splatting a value is not useful.
17143 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
17144 across a vec_duplicate and a paradoxical subreg forming a vector
17145 mode to a vec_concat.
17147 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17149 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
17150 -march=armv8.3-a variants.
17151 * config/arm/t-multilib: Likewise.
17152 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
17154 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17156 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
17158 (cceq_ior_compare_complement): Give it a name so I can use it, and
17159 change boolean_or_operator predicate to boolean_operator so it can
17160 be used to generate a crand.
17161 (eqne): New code iterator.
17162 (bd/bd_neg): New code_attrs.
17163 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
17164 a single define_insn.
17165 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
17166 decrement (bdnzt/bdnzf/bdzt/bdzf).
17167 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
17168 with the new names of the branch decrement patterns, and added the
17169 names of the branch decrement conditional patterns.
17171 2018-01-08 Richard Biener <rguenther@suse.de>
17173 PR tree-optimization/83563
17174 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
17177 2018-01-08 Richard Biener <rguenther@suse.de>
17179 PR middle-end/83713
17180 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
17182 2018-01-08 Richard Biener <rguenther@suse.de>
17184 PR tree-optimization/83685
17185 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
17186 references to abnormals.
17188 2018-01-08 Richard Biener <rguenther@suse.de>
17191 * dwarf2out.c (output_indirect_strings): Handle empty
17192 skeleton_debug_str_hash.
17193 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
17195 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17197 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
17198 (emit_store_direct): Likewise.
17199 (arc_trampoline_adjust_address): Likewise.
17200 (arc_asm_trampoline_template): New function.
17201 (arc_initialize_trampoline): Use asm_trampoline_template.
17202 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
17203 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
17204 * config/arc/arc.md (flush_icache): Delete pattern.
17206 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
17208 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
17209 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
17212 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17215 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
17216 by not USED_FOR_TARGET.
17217 (make_pass_resolve_sw_modes): Likewise.
17219 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
17221 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
17224 2018-01-08 Richard Biener <rguenther@suse.de>
17226 PR middle-end/83580
17227 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
17229 2018-01-08 Richard Biener <rguenther@suse.de>
17231 PR middle-end/83517
17232 * match.pd ((t * 2) / 2) -> t): Add missing :c.
17234 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
17236 PR middle-end/81897
17237 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
17238 basic blocks with a small number of successors.
17239 (convert_control_dep_chain_into_preds): Improve handling of
17241 (dump_predicates): Split apart into...
17242 (dump_pred_chain): ...here...
17243 (dump_pred_info): ...and here.
17244 (can_one_predicate_be_invalidated_p): Add debugging printfs.
17245 (can_chain_union_be_invalidated_p): Improve check for invalidation
17247 (uninit_uses_cannot_happen): Avoid unnecessary if
17248 convert_control_dep_chain_into_preds yielded nothing.
17250 2018-01-06 Martin Sebor <msebor@redhat.com>
17252 PR tree-optimization/83640
17253 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
17254 subtracting negative offset from size.
17255 (builtin_access::overlap): Adjust offset bounds of the access to fall
17256 within the size of the object if possible.
17258 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
17260 PR rtl-optimization/83699
17261 * expmed.c (extract_bit_field_1): Restrict the vector usage of
17262 extract_bit_field_as_subreg to cases in which the extracted
17263 value is also a vector.
17265 * lra-constraints.c (process_alt_operands): Test for the equivalence
17266 substitutions when detecting a possible reload cycle.
17268 2018-01-06 Jakub Jelinek <jakub@redhat.com>
17271 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
17272 by default if flag_selective_schedling{,2}. Formatting fixes.
17274 PR rtl-optimization/83682
17275 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
17276 if it has non-VECTOR_MODE element mode.
17277 (vec_duplicate_p): Likewise.
17279 PR middle-end/83694
17280 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
17281 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
17283 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17286 * config/i386/i386-builtin.def
17287 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
17288 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
17289 Require also OPTION_MASK_ISA_AVX512F in addition to
17290 OPTION_MASK_ISA_GFNI.
17291 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
17292 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
17293 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
17294 to OPTION_MASK_ISA_GFNI.
17295 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
17296 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
17297 OPTION_MASK_ISA_AVX512BW.
17298 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
17299 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
17300 addition to OPTION_MASK_ISA_GFNI.
17301 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
17302 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
17303 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
17304 to OPTION_MASK_ISA_GFNI.
17305 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
17306 a requirement for all ISAs rather than any of them with a few
17308 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
17310 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
17311 bitmasks to be enabled with 3 exceptions, instead of requiring any
17312 enabled ISA with lots of exceptions.
17313 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
17314 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
17315 Change avx512bw in isa attribute to avx512f.
17316 * config/i386/sgxintrin.h: Add license boilerplate.
17317 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
17318 to __AVX512F__ and __AVX512VL to __AVX512VL__.
17319 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
17320 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
17322 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
17323 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
17324 temporarily sse2 rather than sse if not enabled already.
17327 * config/i386/sse.md (VI248_VLBW): Rename to ...
17328 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
17329 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
17330 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
17331 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
17332 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
17333 mode iterator instead of VI248_VLBW.
17335 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
17337 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
17338 (record_modified): Skip clobbers; add debug output.
17339 (param_change_prob): Use sreal frequencies.
17341 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17343 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
17344 punt for user-aligned variables.
17346 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
17348 * tree-chrec.c (chrec_contains_symbols): Return true for
17351 2018-01-05 Sudakshina Das <sudi.das@arm.com>
17354 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
17355 of (x|y) == x for BICS pattern.
17357 2018-01-05 Jakub Jelinek <jakub@redhat.com>
17359 PR tree-optimization/83605
17360 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
17361 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
17364 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
17366 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
17367 * config/epiphany/rtems.h: New file.
17369 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17370 Uros Bizjak <ubizjak@gmail.com>
17373 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
17374 QIreg_operand instead of register_operand predicate.
17375 * config/i386/i386.c (ix86_rop_should_change_byte_p,
17376 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
17377 comments instead of -fmitigate[-_]rop.
17379 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17382 * cgraphunit.c (symbol_table::compile): Switch to text_section
17383 before calling assembly_start debug hook.
17384 * run-rtl-passes.c (run_rtl_passes): Likewise.
17387 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17389 * tree-vrp.c (extract_range_from_binary_expr_1): Check
17390 range_int_cst_p rather than !symbolic_range_p before calling
17391 extract_range_from_multiplicative_op_1.
17393 2018-01-04 Jeff Law <law@redhat.com>
17395 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
17396 redundant test in assertion.
17398 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17400 * doc/rtl.texi: Document machine_mode wrapper classes.
17402 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17404 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
17405 using tree_to_uhwi.
17407 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17409 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
17410 the VEC_PERM_EXPR fold to fail.
17412 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17415 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
17416 to switched_sections.
17418 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17421 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
17422 test for d.testing.
17424 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
17427 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
17428 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
17430 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17433 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
17434 is BLKmode and bitpos not zero or mode change is needed.
17436 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
17439 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
17442 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
17445 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
17446 instead of MULT rtx. Update all corresponding splitters.
17447 (*saddl_se): Ditto.
17448 (*ssub<modesuffix>): Ditto.
17449 (*ssubl_se): Ditto.
17450 (*cmp_sadd_di): Update split patterns.
17451 (*cmp_sadd_si): Ditto.
17452 (*cmp_sadd_sidi): Ditto.
17453 (*cmp_ssub_di): Ditto.
17454 (*cmp_ssub_si): Ditto.
17455 (*cmp_ssub_sidi): Ditto.
17456 * config/alpha/predicates.md (const23_operand): New predicate.
17457 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
17458 Look for ASHIFT, not MULT inner operand.
17459 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
17461 2018-01-04 Martin Liska <mliska@suse.cz>
17463 PR gcov-profile/83669
17464 * gcov.c (output_intermediate_file): Add version to intermediate
17466 * doc/gcov.texi: Document new field 'version' in intermediate
17467 file format. Fix location of '-k' option of gcov command.
17469 2018-01-04 Martin Liska <mliska@suse.cz>
17472 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
17474 2018-01-04 Jakub Jelinek <jakub@redhat.com>
17476 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
17478 2018-01-03 Martin Sebor <msebor@redhat.com>
17480 PR tree-optimization/83655
17481 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
17482 checking calls with invalid arguments.
17484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17486 * tree-vect-stmts.c (vect_get_store_rhs): New function.
17487 (vectorizable_mask_load_store): Delete.
17488 (vectorizable_call): Return false for masked loads and stores.
17489 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
17490 instead of gimple_assign_rhs1.
17491 (vectorizable_load): Handle IFN_MASK_LOAD.
17492 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
17494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17496 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
17498 (vectorizable_mask_load_store): ...here.
17499 (vectorizable_load): ...and here.
17501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17503 * tree-vect-stmts.c (vect_build_all_ones_mask)
17504 (vect_build_zero_merge_argument): New functions, split out from...
17505 (vectorizable_load): ...here.
17507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17509 * tree-vect-stmts.c (vect_check_store_rhs): New function,
17511 (vectorizable_mask_load_store): ...here.
17512 (vectorizable_store): ...and here.
17514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17516 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
17518 (vectorizable_mask_load_store): ...here.
17520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17522 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
17523 (vect_model_store_cost): Take a vec_load_store_type instead of a
17525 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
17526 (vect_model_store_cost): Take a vec_load_store_type instead of a
17528 (vectorizable_mask_load_store): Update accordingly.
17529 (vectorizable_store): Likewise.
17530 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
17532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17534 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
17535 IFN_MASK_LOAD calls here rather than...
17536 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
17538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17539 Alan Hayward <alan.hayward@arm.com>
17540 David Sherwood <david.sherwood@arm.com>
17542 * expmed.c (extract_bit_field_1): For vector extracts,
17543 fall back to extract_bit_field_as_subreg if vec_extract
17546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17547 Alan Hayward <alan.hayward@arm.com>
17548 David Sherwood <david.sherwood@arm.com>
17550 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
17551 they are variable or constant sized.
17552 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
17553 slots for constant-sized data.
17555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17556 Alan Hayward <alan.hayward@arm.com>
17557 David Sherwood <david.sherwood@arm.com>
17559 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
17560 handling COND_EXPRs with boolean comparisons, try to find a better
17561 basis for the mask type than the boolean itself.
17563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17565 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
17566 is calculated and how it can be overridden.
17567 * genmodes.c (max_bitsize_mode_any_mode): New variable.
17568 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
17570 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
17573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17574 Alan Hayward <alan.hayward@arm.com>
17575 David Sherwood <david.sherwood@arm.com>
17577 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
17578 Remove the mode argument.
17579 (aarch64_simd_valid_immediate): Remove the mode and inverse
17581 * config/aarch64/iterators.md (bitsize): New iterator.
17582 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
17583 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
17584 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
17585 aarch64_simd_valid_immediate.
17586 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
17587 (aarch64_reg_or_bic_imm): Likewise.
17588 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
17589 with an insn_type enum and msl with a modifier_type enum.
17590 Replace element_width with a scalar_mode. Change the shift
17591 to unsigned int. Add constructors for scalar_float_mode and
17592 scalar_int_mode elements.
17593 (aarch64_vect_float_const_representable_p): Delete.
17594 (aarch64_can_const_movi_rtx_p)
17595 (aarch64_simd_scalar_immediate_valid_for_move)
17596 (aarch64_simd_make_constant): Update call to
17597 aarch64_simd_valid_immediate.
17598 (aarch64_advsimd_valid_immediate_hs): New function.
17599 (aarch64_advsimd_valid_immediate): Likewise.
17600 (aarch64_simd_valid_immediate): Remove mode and inverse
17601 arguments. Rewrite to use the above. Use const_vec_duplicate_p
17602 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
17603 and aarch64_float_const_representable_p on the result.
17604 (aarch64_output_simd_mov_immediate): Remove mode argument.
17605 Update call to aarch64_simd_valid_immediate and use of
17606 simd_immediate_info.
17607 (aarch64_output_scalar_simd_mov_immediate): Update call
17610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17611 Alan Hayward <alan.hayward@arm.com>
17612 David Sherwood <david.sherwood@arm.com>
17614 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
17615 (mode_nunits): Likewise CONST_MODE_NUNITS.
17616 * machmode.def (ADJUST_NUNITS): Document.
17617 * genmodes.c (mode_data::need_nunits_adj): New field.
17618 (blank_mode): Update accordingly.
17619 (adj_nunits): New variable.
17620 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
17622 (emit_mode_size_inline): Set need_bytesize_adj for all modes
17623 listed in adj_nunits.
17624 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
17625 listed in adj_nunits. Don't emit case statements for such modes.
17626 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
17627 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
17628 nothing if adj_nunits is nonnull.
17629 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
17630 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
17631 (emit_mode_fbit): Update use of print_maybe_const_decl.
17632 (emit_move_size): Likewise. Treat the array as non-const
17634 (emit_mode_adjustments): Handle adj_nunits.
17636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17638 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
17639 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
17640 (VECTOR_MODES): Use it.
17641 (make_vector_modes): Take the prefix as an argument.
17643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17644 Alan Hayward <alan.hayward@arm.com>
17645 David Sherwood <david.sherwood@arm.com>
17647 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
17648 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
17649 for MODE_VECTOR_BOOL.
17650 * machmode.def (VECTOR_BOOL_MODE): Document.
17651 * genmodes.c (VECTOR_BOOL_MODE): New macro.
17652 (make_vector_bool_mode): New function.
17653 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
17655 * lto-streamer-in.c (lto_input_mode_table): Likewise.
17656 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
17658 * stor-layout.c (int_mode_for_mode): Likewise.
17659 * tree.c (build_vector_type_for_mode): Likewise.
17660 * varasm.c (output_constant_pool_2): Likewise.
17661 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
17662 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
17663 for MODE_VECTOR_BOOL.
17664 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
17665 of mode class checks.
17666 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
17667 instead of a list of mode class checks.
17668 (expand_vector_scalar_condition): Likewise.
17669 (type_for_widest_vector_mode): Handle BImode as an inner mode.
17671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17672 Alan Hayward <alan.hayward@arm.com>
17673 David Sherwood <david.sherwood@arm.com>
17675 * machmode.h (mode_size): Change from unsigned short to
17677 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
17678 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17679 or if measurement_type is not polynomial.
17680 (fixed_size_mode::includes_p): Check for constant-sized modes.
17681 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
17682 return a poly_uint16 rather than an unsigned short.
17683 (emit_mode_size): Change the type of mode_size from unsigned short
17684 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
17685 (emit_mode_adjustments): Cope with polynomial vector sizes.
17686 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17688 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17690 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
17691 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
17692 * caller-save.c (setup_save_areas): Likewise.
17693 (replace_reg_with_saved_mem): Likewise.
17694 * calls.c (emit_library_call_value_1): Likewise.
17695 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
17696 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
17697 (gen_lowpart_for_combine): Likewise.
17698 * convert.c (convert_to_integer_1): Likewise.
17699 * cse.c (equiv_constant, cse_insn): Likewise.
17700 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
17701 (cselib_subst_to_values): Likewise.
17702 * dce.c (word_dce_process_block): Likewise.
17703 * df-problems.c (df_word_lr_mark_ref): Likewise.
17704 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
17705 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
17706 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
17707 (rtl_for_decl_location): Likewise.
17708 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
17709 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
17710 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
17711 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
17712 (expand_expr_real_1): Likewise.
17713 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
17714 (pad_below): Likewise.
17715 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17716 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
17717 * ira.c (get_subreg_tracking_sizes): Likewise.
17718 * ira-build.c (ira_create_allocno_objects): Likewise.
17719 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
17720 (ira_sort_regnos_for_alter_reg): Likewise.
17721 * ira-costs.c (record_operand_costs): Likewise.
17722 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
17723 (resolve_simple_move): Likewise.
17724 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
17725 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
17726 (lra_constraints): Likewise.
17727 (CONST_POOL_OK_P): Reject variable-sized modes.
17728 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
17729 (add_pseudo_to_slot, lra_spill): Likewise.
17730 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17731 * optabs-query.c (get_best_extraction_insn): Likewise.
17732 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17733 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
17734 (expand_mult_highpart, valid_multiword_target_p): Likewise.
17735 * recog.c (offsettable_address_addr_space_p): Likewise.
17736 * regcprop.c (maybe_mode_change): Likewise.
17737 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
17738 * regrename.c (build_def_use): Likewise.
17739 * regstat.c (dump_reg_info): Likewise.
17740 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
17741 (find_reloads, find_reloads_subreg_address): Likewise.
17742 * reload1.c (eliminate_regs_1): Likewise.
17743 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
17744 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
17745 (simplify_binary_operation_1, simplify_subreg): Likewise.
17746 * targhooks.c (default_function_arg_padding): Likewise.
17747 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
17748 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
17749 (verify_gimple_assign_ternary): Likewise.
17750 * tree-inline.c (estimate_move_cost): Likewise.
17751 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17752 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
17753 (get_address_cost_ainc): Likewise.
17754 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
17755 (vect_supportable_dr_alignment): Likewise.
17756 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17757 (vectorizable_reduction): Likewise.
17758 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
17759 (vectorizable_operation, vectorizable_load): Likewise.
17760 * tree.c (build_same_sized_truth_vector_type): Likewise.
17761 * valtrack.c (cleanup_auto_inc_dec): Likewise.
17762 * var-tracking.c (emit_note_insn_var_location): Likewise.
17763 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
17764 (ADDR_VEC_ALIGN): Likewise.
17766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17767 Alan Hayward <alan.hayward@arm.com>
17768 David Sherwood <david.sherwood@arm.com>
17770 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
17772 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
17773 or if measurement_type is polynomial.
17774 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
17775 * combine.c (make_extraction): Likewise.
17776 * dse.c (find_shift_sequence): Likewise.
17777 * dwarf2out.c (mem_loc_descriptor): Likewise.
17778 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
17779 (extract_bit_field, extract_low_bits): Likewise.
17780 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
17781 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
17782 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
17783 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
17784 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
17785 * reload.c (find_reloads): Likewise.
17786 * reload1.c (alter_reg): Likewise.
17787 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
17788 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
17789 * tree-if-conv.c (predicate_mem_writes): Likewise.
17790 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
17791 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
17792 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
17793 * valtrack.c (dead_debug_insert_temp): Likewise.
17794 * varasm.c (mergeable_constant_section): Likewise.
17795 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
17797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17798 Alan Hayward <alan.hayward@arm.com>
17799 David Sherwood <david.sherwood@arm.com>
17801 * expr.c (expand_assignment): Cope with polynomial mode sizes
17802 when assigning to a CONCAT.
17804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17805 Alan Hayward <alan.hayward@arm.com>
17806 David Sherwood <david.sherwood@arm.com>
17808 * machmode.h (mode_precision): Change from unsigned short to
17810 (mode_to_precision): Return a poly_uint16 rather than an unsigned
17812 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
17813 or if measurement_type is not polynomial.
17814 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
17815 in which the mode is already known to be a scalar_int_mode.
17816 * genmodes.c (emit_mode_precision): Change the type of mode_precision
17817 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
17819 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17820 for GET_MODE_PRECISION.
17821 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17822 for GET_MODE_PRECISION.
17823 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
17825 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
17826 (expand_field_assignment, make_extraction): Likewise.
17827 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
17828 (get_last_value): Likewise.
17829 * convert.c (convert_to_integer_1): Likewise.
17830 * cse.c (cse_insn): Likewise.
17831 * expr.c (expand_expr_real_1): Likewise.
17832 * lra-constraints.c (simplify_operand_subreg): Likewise.
17833 * optabs-query.c (can_atomic_load_p): Likewise.
17834 * optabs.c (expand_atomic_load): Likewise.
17835 (expand_atomic_store): Likewise.
17836 * ree.c (combine_reaching_defs): Likewise.
17837 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
17838 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
17839 * tree.h (type_has_mode_precision_p): Likewise.
17840 * ubsan.c (instrument_si_overflow): Likewise.
17842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17843 Alan Hayward <alan.hayward@arm.com>
17844 David Sherwood <david.sherwood@arm.com>
17846 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
17847 polynomial numbers of units.
17848 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
17849 (valid_vector_subparts_p): New function.
17850 (build_vector_type): Remove temporary shim and take the number
17851 of units as a poly_uint64 rather than an int.
17852 (build_opaque_vector_type): Take the number of units as a
17853 poly_uint64 rather than an int.
17854 * tree.c (build_vector_from_ctor): Handle polynomial
17855 TYPE_VECTOR_SUBPARTS.
17856 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
17857 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
17858 (build_vector_from_val): If the number of units is variable,
17859 use build_vec_duplicate_cst for constant operands and
17860 VEC_DUPLICATE_EXPR otherwise.
17861 (make_vector_type): Remove temporary is_constant ().
17862 (build_vector_type, build_opaque_vector_type): Take the number of
17863 units as a poly_uint64 rather than an int.
17864 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
17866 * cfgexpand.c (expand_debug_expr): Likewise.
17867 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
17868 (store_constructor, expand_expr_real_1): Likewise.
17869 (const_scalar_mask_from_tree): Likewise.
17870 * fold-const-call.c (fold_const_reduction): Likewise.
17871 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
17872 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
17873 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
17874 (fold_relational_const): Likewise.
17875 (native_interpret_vector): Likewise. Change the size from an
17876 int to an unsigned int.
17877 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
17878 TYPE_VECTOR_SUBPARTS.
17879 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
17880 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
17881 duplicating a non-constant operand into a variable-length vector.
17882 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
17883 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
17884 * ipa-icf.c (sem_variable::equals): Likewise.
17885 * match.pd: Likewise.
17886 * omp-simd-clone.c (simd_clone_subparts): Likewise.
17887 * print-tree.c (print_node): Likewise.
17888 * stor-layout.c (layout_type): Likewise.
17889 * targhooks.c (default_builtin_vectorization_cost): Likewise.
17890 * tree-cfg.c (verify_gimple_comparison): Likewise.
17891 (verify_gimple_assign_binary): Likewise.
17892 (verify_gimple_assign_ternary): Likewise.
17893 (verify_gimple_assign_single): Likewise.
17894 * tree-pretty-print.c (dump_generic_node): Likewise.
17895 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17896 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
17897 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
17898 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
17899 (vect_shift_permute_load_chain): Likewise.
17900 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
17901 (expand_vector_condition, optimize_vector_constructor): Likewise.
17902 (lower_vec_perm, get_compute_type): Likewise.
17903 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17904 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
17905 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
17906 (vect_recog_mask_conversion_pattern): Likewise.
17907 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
17908 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
17909 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17910 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
17911 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
17912 (vectorizable_shift, vectorizable_operation, vectorizable_store)
17913 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
17914 (supportable_widening_operation): Likewise.
17915 (supportable_narrowing_operation): Likewise.
17916 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
17918 * varasm.c (output_constant): Likewise.
17920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17921 Alan Hayward <alan.hayward@arm.com>
17922 David Sherwood <david.sherwood@arm.com>
17924 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
17925 so that both the length == 3 and length != 3 cases set up their
17926 own permute vectors. Add comments explaining why we know the
17927 number of elements is constant.
17928 (vect_permute_load_chain): Likewise.
17930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17931 Alan Hayward <alan.hayward@arm.com>
17932 David Sherwood <david.sherwood@arm.com>
17934 * machmode.h (mode_nunits): Change from unsigned char to
17936 (ONLY_FIXED_SIZE_MODES): New macro.
17937 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
17938 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
17939 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
17941 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
17942 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
17943 or if measurement_type is not polynomial.
17944 * genmodes.c (ZERO_COEFFS): New macro.
17945 (emit_mode_nunits_inline): Make mode_nunits_inline return a
17947 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
17948 Use ZERO_COEFFS when emitting initializers.
17949 * data-streamer.h (bp_pack_poly_value): New function.
17950 (bp_unpack_poly_value): Likewise.
17951 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
17952 for GET_MODE_NUNITS.
17953 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
17954 for GET_MODE_NUNITS.
17955 * tree.c (make_vector_type): Remove temporary shim and make
17956 the real function take the number of units as a poly_uint64
17957 rather than an int.
17958 (build_vector_type_for_mode): Handle polynomial nunits.
17959 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
17960 * emit-rtl.c (const_vec_series_p_1): Likewise.
17961 (gen_rtx_CONST_VECTOR): Likewise.
17962 * fold-const.c (test_vec_duplicate_folding): Likewise.
17963 * genrecog.c (validate_pattern): Likewise.
17964 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
17965 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
17966 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
17967 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
17968 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
17969 * rtlanal.c (subreg_get_info): Likewise.
17970 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17971 (vect_grouped_load_supported): Likewise.
17972 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
17973 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
17974 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
17975 (simplify_const_unary_operation, simplify_binary_operation_1)
17976 (simplify_const_binary_operation, simplify_ternary_operation)
17977 (test_vector_ops_duplicate, test_vector_ops): Likewise.
17978 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
17979 instead of CONST_VECTOR_NUNITS.
17980 * varasm.c (output_constant_pool_2): Likewise.
17981 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
17982 explicit-encoded elements in the XVEC for variable-length vectors.
17984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17986 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
17988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17989 Alan Hayward <alan.hayward@arm.com>
17990 David Sherwood <david.sherwood@arm.com>
17992 * coretypes.h (fixed_size_mode): Declare.
17993 (fixed_size_mode_pod): New typedef.
17994 * builtins.h (target_builtins::x_apply_args_mode)
17995 (target_builtins::x_apply_result_mode): Change type to
17996 fixed_size_mode_pod.
17997 * builtins.c (apply_args_size, apply_result_size, result_vector)
17998 (expand_builtin_apply_args_1, expand_builtin_apply)
17999 (expand_builtin_return): Update accordingly.
18001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18003 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18004 * cselib.c (cselib_hash_rtx): Likewise.
18005 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18006 CONST_VECTOR encoding.
18008 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18009 Jeff Law <law@redhat.com>
18012 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18013 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18014 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18015 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18018 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18019 explicitly probe *sp in a noreturn function if there were any callee
18020 register saves or frame pointer is needed.
18022 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18025 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18026 BLKmode for ternary, binary or unary expressions.
18029 * var-tracking.c (delete_vta_debug_insn): New inline function.
18030 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18031 insns from get_insns () to NULL instead of each bb separately.
18032 Use delete_vta_debug_insn. No longer static.
18033 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18034 delete_vta_debug_insns callers.
18035 * rtl.h (delete_vta_debug_insns): Declare.
18036 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18037 instead of variable_tracking_main.
18039 2018-01-03 Martin Sebor <msebor@redhat.com>
18041 PR tree-optimization/83603
18042 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18043 arguments past the endof the argument list in functions declared
18044 without a prototype.
18045 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18046 Avoid checking when arguments are null.
18048 2018-01-03 Martin Sebor <msebor@redhat.com>
18051 * doc/extend.texi (attribute const): Fix a typo.
18052 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18053 issuing -Wsuggest-attribute for void functions.
18055 2018-01-03 Martin Sebor <msebor@redhat.com>
18057 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18058 offset_int::from instead of wide_int::to_shwi.
18059 (maybe_diag_overlap): Remove assertion.
18060 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18061 * gimple-ssa-sprintf.c (format_directive): Same.
18062 (parse_directive): Same.
18063 (sprintf_dom_walker::compute_format_length): Same.
18064 (try_substitute_return_value): Same.
18066 2018-01-03 Jeff Law <law@redhat.com>
18068 PR middle-end/83654
18069 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
18070 non-constant residual for zero at runtime and avoid probing in
18071 that case. Reorganize code for trailing problem to mirror handling
18074 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18076 PR tree-optimization/83501
18077 * tree-ssa-strlen.c (get_string_cst): New.
18078 (handle_char_store): Call get_string_cst.
18080 2018-01-03 Martin Liska <mliska@suse.cz>
18082 PR tree-optimization/83593
18083 * tree-ssa-strlen.c: Include tree-cfg.h.
18084 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
18085 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
18086 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
18088 (strlen_dom_walker::before_dom_children): Call
18089 gimple_purge_dead_eh_edges. Dump tranformation with details
18091 (strlen_dom_walker::before_dom_children): Update call by adding
18092 new argument cleanup_eh.
18093 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
18095 2018-01-03 Martin Liska <mliska@suse.cz>
18098 * cif-code.def (VARIADIC_THUNK): New enum value.
18099 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
18102 2018-01-03 Jan Beulich <jbeulich@suse.com>
18104 * sse.md (mov<mode>_internal): Tighten condition for when to use
18105 vmovdqu<ssescalarsize> for TI and OI modes.
18107 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18109 Update copyright years.
18111 2018-01-03 Martin Liska <mliska@suse.cz>
18114 * ipa-visibility.c (function_and_variable_visibility): Skip
18115 functions with noipa attribure.
18117 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18119 * gcc.c (process_command): Update copyright notice dates.
18120 * gcov-dump.c (print_version): Ditto.
18121 * gcov.c (print_version): Ditto.
18122 * gcov-tool.c (print_version): Ditto.
18123 * gengtype.c (create_file): Ditto.
18124 * doc/cpp.texi: Bump @copying's copyright year.
18125 * doc/cppinternals.texi: Ditto.
18126 * doc/gcc.texi: Ditto.
18127 * doc/gccint.texi: Ditto.
18128 * doc/gcov.texi: Ditto.
18129 * doc/install.texi: Ditto.
18130 * doc/invoke.texi: Ditto.
18132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18134 * vector-builder.h (vector_builder::m_full_nelts): Change from
18135 unsigned int to poly_uint64.
18136 (vector_builder::full_nelts): Update prototype accordingly.
18137 (vector_builder::new_vector): Likewise.
18138 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
18139 (vector_builder::operator ==): Likewise.
18140 (vector_builder::finalize): Likewise.
18141 * int-vector-builder.h (int_vector_builder::int_vector_builder):
18142 Take the number of elements as a poly_uint64 rather than an
18144 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
18145 from unsigned int to poly_uint64.
18146 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
18147 (vec_perm_indices::new_vector): Likewise.
18148 (vec_perm_indices::length): Likewise.
18149 (vec_perm_indices::nelts_per_input): Likewise.
18150 (vec_perm_indices::input_nelts): Likewise.
18151 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
18152 number of elements per input as a poly_uint64 rather than an
18153 unsigned int. Use the original encoding for variable-length
18154 vectors, rather than clamping each individual element.
18155 For the second and subsequent elements in each pattern,
18156 clamp the step and base before clamping their sum.
18157 (vec_perm_indices::series_p): Handle polynomial element counts.
18158 (vec_perm_indices::all_in_range_p): Likewise.
18159 (vec_perm_indices_to_tree): Likewise.
18160 (vec_perm_indices_to_rtx): Likewise.
18161 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
18162 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
18163 (tree_vector_builder::new_binary_operation): Handle polynomial
18164 element counts. Return false if we need to know the number
18165 of elements at compile time.
18166 * fold-const.c (fold_vec_perm): Punt if the number of elements
18167 isn't known at compile time.
18169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18171 * vec-perm-indices.h (vec_perm_builder): Change element type
18172 from HOST_WIDE_INT to poly_int64.
18173 (vec_perm_indices::element_type): Update accordingly.
18174 (vec_perm_indices::clamp): Handle polynomial element_types.
18175 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18176 (vec_perm_indices::all_in_range_p): Likewise.
18177 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
18179 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
18180 polynomial vec_perm_indices element types.
18181 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
18182 * fold-const.c (fold_vec_perm): Likewise.
18183 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
18184 * tree-vect-generic.c (lower_vec_perm): Likewise.
18185 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18186 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
18187 element type to HOST_WIDE_INT.
18189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18190 Alan Hayward <alan.hayward@arm.com>
18191 David Sherwood <david.sherwood@arm.com>
18193 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
18194 rather than an int. Use plus_constant.
18195 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
18196 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
18198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18199 Alan Hayward <alan.hayward@arm.com>
18200 David Sherwood <david.sherwood@arm.com>
18202 * calls.c (emit_call_1, expand_call): Change struct_value_size from
18203 a HOST_WIDE_INT to a poly_int64.
18205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18206 Alan Hayward <alan.hayward@arm.com>
18207 David Sherwood <david.sherwood@arm.com>
18209 * calls.c (load_register_parameters): Cope with polynomial
18210 mode sizes. Require a constant size for BLKmode parameters
18211 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
18212 forces a parameter to be padded at the lsb end in order to
18213 fill a complete number of words, require the parameter size
18214 to be ordered wrt UNITS_PER_WORD.
18216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18217 Alan Hayward <alan.hayward@arm.com>
18218 David Sherwood <david.sherwood@arm.com>
18220 * reload1.c (spill_stack_slot_width): Change element type
18221 from unsigned int to poly_uint64_pod.
18222 (alter_reg): Treat mode sizes as polynomial.
18224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18225 Alan Hayward <alan.hayward@arm.com>
18226 David Sherwood <david.sherwood@arm.com>
18228 * reload.c (complex_word_subreg_p): New function.
18229 (reload_inner_reg_of_subreg, push_reload): Use it.
18231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18232 Alan Hayward <alan.hayward@arm.com>
18233 David Sherwood <david.sherwood@arm.com>
18235 * lra-constraints.c (process_alt_operands): Reject matched
18236 operands whose sizes aren't ordered.
18237 (match_reload): Refer to this check here.
18239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18240 Alan Hayward <alan.hayward@arm.com>
18241 David Sherwood <david.sherwood@arm.com>
18243 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
18244 that the mode size is in the set {1, 2, 4, 8, 16}.
18246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18247 Alan Hayward <alan.hayward@arm.com>
18248 David Sherwood <david.sherwood@arm.com>
18250 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
18251 Use plus_constant instead of gen_rtx_PLUS.
18253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18254 Alan Hayward <alan.hayward@arm.com>
18255 David Sherwood <david.sherwood@arm.com>
18257 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
18258 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
18259 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
18260 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
18261 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
18262 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
18263 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
18264 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
18265 * config/i386/i386.c (ix86_push_rounding): ...this new function.
18266 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
18268 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
18269 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
18270 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
18271 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
18272 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
18273 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
18274 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
18275 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
18276 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
18277 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
18279 * expr.c (emit_move_resolve_push): Treat the input and result
18280 of PUSH_ROUNDING as a poly_int64.
18281 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
18282 (emit_push_insn): Likewise.
18283 * lra-eliminations.c (mark_not_eliminable): Likewise.
18284 * recog.c (push_operand): Likewise.
18285 * reload1.c (elimination_effects): Likewise.
18286 * rtlanal.c (nonzero_bits1): Likewise.
18287 * calls.c (store_one_arg): Likewise. Require the padding to be
18288 known at compile time.
18290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18291 Alan Hayward <alan.hayward@arm.com>
18292 David Sherwood <david.sherwood@arm.com>
18294 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
18295 Use plus_constant instead of gen_rtx_PLUS.
18297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18298 Alan Hayward <alan.hayward@arm.com>
18299 David Sherwood <david.sherwood@arm.com>
18301 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
18302 rather than an int.
18304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18305 Alan Hayward <alan.hayward@arm.com>
18306 David Sherwood <david.sherwood@arm.com>
18308 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
18309 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
18310 via stack temporaries. Treat the mode size as polynomial too.
18312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18313 Alan Hayward <alan.hayward@arm.com>
18314 David Sherwood <david.sherwood@arm.com>
18316 * expr.c (expand_expr_real_2): When handling conversions involving
18317 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
18318 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
18319 as a poly_uint64 too.
18321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18322 Alan Hayward <alan.hayward@arm.com>
18323 David Sherwood <david.sherwood@arm.com>
18325 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
18327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18328 Alan Hayward <alan.hayward@arm.com>
18329 David Sherwood <david.sherwood@arm.com>
18331 * combine.c (can_change_dest_mode): Handle polynomial
18332 REGMODE_NATURAL_SIZE.
18333 * expmed.c (store_bit_field_1): Likewise.
18334 * expr.c (store_constructor): Likewise.
18335 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
18336 and polynomial REGMODE_NATURAL_SIZE.
18337 (gen_lowpart_common): Likewise.
18338 * reginfo.c (record_subregs_of_mode): Likewise.
18339 * rtlanal.c (read_modify_subreg_p): Likewise.
18341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18342 Alan Hayward <alan.hayward@arm.com>
18343 David Sherwood <david.sherwood@arm.com>
18345 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
18346 numbers of elements.
18348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18349 Alan Hayward <alan.hayward@arm.com>
18350 David Sherwood <david.sherwood@arm.com>
18352 * match.pd: Cope with polynomial numbers of vector elements.
18354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18355 Alan Hayward <alan.hayward@arm.com>
18356 David Sherwood <david.sherwood@arm.com>
18358 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
18359 in a POINTER_PLUS_EXPR.
18361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18362 Alan Hayward <alan.hayward@arm.com>
18363 David Sherwood <david.sherwood@arm.com>
18365 * omp-simd-clone.c (simd_clone_subparts): New function.
18366 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
18367 (ipa_simd_modify_function_body): Likewise.
18369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18370 Alan Hayward <alan.hayward@arm.com>
18371 David Sherwood <david.sherwood@arm.com>
18373 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
18374 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
18375 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
18376 (expand_vector_condition, vector_element): Likewise.
18377 (subparts_gt): New function.
18378 (get_compute_type): Use subparts_gt.
18379 (count_type_subparts): Delete.
18380 (expand_vector_operations_1): Use subparts_gt instead of
18381 count_type_subparts.
18383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18384 Alan Hayward <alan.hayward@arm.com>
18385 David Sherwood <david.sherwood@arm.com>
18387 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
18388 (vect_compile_time_alias): ...this new function. Do the calculation
18389 on poly_ints rather than trees.
18390 (vect_prune_runtime_alias_test_list): Update call accordingly.
18392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18393 Alan Hayward <alan.hayward@arm.com>
18394 David Sherwood <david.sherwood@arm.com>
18396 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
18398 (vect_schedule_slp_instance): Likewise.
18400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18401 Alan Hayward <alan.hayward@arm.com>
18402 David Sherwood <david.sherwood@arm.com>
18404 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
18405 constant and extern definitions for variable-length vectors.
18406 (vect_get_constant_vectors): Note that the number of units
18407 is known to be constant.
18409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18410 Alan Hayward <alan.hayward@arm.com>
18411 David Sherwood <david.sherwood@arm.com>
18413 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
18414 of units as polynomial. Choose between WIDE and NARROW based
18417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18418 Alan Hayward <alan.hayward@arm.com>
18419 David Sherwood <david.sherwood@arm.com>
18421 * tree-vect-stmts.c (simd_clone_subparts): New function.
18422 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
18424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18425 Alan Hayward <alan.hayward@arm.com>
18426 David Sherwood <david.sherwood@arm.com>
18428 * tree-vect-stmts.c (vectorizable_call): Treat the number of
18429 vectors as polynomial. Use build_index_vector for
18430 IFN_GOMP_SIMD_LANE.
18432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18433 Alan Hayward <alan.hayward@arm.com>
18434 David Sherwood <david.sherwood@arm.com>
18436 * tree-vect-stmts.c (get_load_store_type): Treat the number of
18437 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
18438 for variable-length vectors.
18439 (vectorizable_mask_load_store): Treat the number of units as
18440 polynomial, asserting that it is constant if the condition has
18441 already been enforced.
18442 (vectorizable_store, vectorizable_load): Likewise.
18444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18445 Alan Hayward <alan.hayward@arm.com>
18446 David Sherwood <david.sherwood@arm.com>
18448 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
18449 of units as polynomial. Punt if we can't tell at compile time
18450 which vector contains the final result.
18452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18453 Alan Hayward <alan.hayward@arm.com>
18454 David Sherwood <david.sherwood@arm.com>
18456 * tree-vect-loop.c (vectorizable_induction): Treat the number
18457 of units as polynomial. Punt on SLP inductions. Use an integer
18458 VEC_SERIES_EXPR for variable-length integer reductions. Use a
18459 cast of such a series for variable-length floating-point
18462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18463 Alan Hayward <alan.hayward@arm.com>
18464 David Sherwood <david.sherwood@arm.com>
18466 * tree.h (build_index_vector): Declare.
18467 * tree.c (build_index_vector): New function.
18468 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
18469 of units as polynomial, forcibly converting it to a constant if
18470 vectorizable_reduction has already enforced the condition.
18471 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
18472 to create a {1,2,3,...} vector.
18473 (vectorizable_reduction): Treat the number of units as polynomial.
18474 Choose vectype_in based on the largest scalar element size rather
18475 than the smallest number of units. Enforce the restrictions
18478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18479 Alan Hayward <alan.hayward@arm.com>
18480 David Sherwood <david.sherwood@arm.com>
18482 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
18483 number of units as polynomial.
18485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18486 Alan Hayward <alan.hayward@arm.com>
18487 David Sherwood <david.sherwood@arm.com>
18489 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
18490 * target.def (autovectorize_vector_sizes): Return the vector sizes
18491 by pointer, using vector_sizes rather than a bitmask.
18492 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
18493 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
18494 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
18496 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
18497 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
18498 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
18499 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
18500 * omp-general.c (omp_max_vf): Likewise.
18501 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18502 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
18503 * tree-vect-loop.c (vect_analyze_loop): Likewise.
18504 * tree-vect-slp.c (vect_slp_bb): Likewise.
18505 * doc/tm.texi: Regenerate.
18506 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
18508 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
18509 the vector size as a poly_uint64 rather than an unsigned int.
18510 (current_vector_size): Change from an unsigned int to a poly_uint64.
18511 (get_vectype_for_scalar_type): Update accordingly.
18512 * tree.h (build_truth_vector_type): Take the size and number of
18513 units as a poly_uint64 rather than an unsigned int.
18514 (build_vector_type): Add a temporary overload that takes
18515 the number of units as a poly_uint64 rather than an unsigned int.
18516 * tree.c (make_vector_type): Likewise.
18517 (build_truth_vector_type): Take the number of units as a poly_uint64
18518 rather than an unsigned int.
18520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18521 Alan Hayward <alan.hayward@arm.com>
18522 David Sherwood <david.sherwood@arm.com>
18524 * target.def (get_mask_mode): Take the number of units and length
18525 as poly_uint64s rather than unsigned ints.
18526 * targhooks.h (default_get_mask_mode): Update accordingly.
18527 * targhooks.c (default_get_mask_mode): Likewise.
18528 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
18529 * doc/tm.texi: Regenerate.
18531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18532 Alan Hayward <alan.hayward@arm.com>
18533 David Sherwood <david.sherwood@arm.com>
18535 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
18536 * omp-general.c (omp_max_vf): Likewise.
18537 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
18538 (expand_omp_simd): Handle polynomial safelen.
18539 * omp-low.c (omplow_simd_context): Add a default constructor.
18540 (omplow_simd_context::max_vf): Change from int to poly_uint64.
18541 (lower_rec_simd_input_clauses): Update accordingly.
18542 (lower_rec_input_clauses): Likewise.
18544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18545 Alan Hayward <alan.hayward@arm.com>
18546 David Sherwood <david.sherwood@arm.com>
18548 * tree-vectorizer.h (vect_nunits_for_cost): New function.
18549 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
18550 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
18551 (vect_analyze_slp_cost): Likewise.
18552 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
18553 (vect_model_load_cost): Likewise.
18555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18556 Alan Hayward <alan.hayward@arm.com>
18557 David Sherwood <david.sherwood@arm.com>
18559 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
18560 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
18561 from an unsigned int * to a poly_uint64_pod *.
18562 (calculate_unrolling_factor): New function.
18563 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
18565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18566 Alan Hayward <alan.hayward@arm.com>
18567 David Sherwood <david.sherwood@arm.com>
18569 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
18570 from an unsigned int to a poly_uint64.
18571 (_loop_vec_info::slp_unrolling_factor): Likewise.
18572 (_loop_vec_info::vectorization_factor): Change from an int
18574 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
18575 (vect_get_num_vectors): New function.
18576 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
18577 (vect_get_num_copies): Use vect_get_num_vectors.
18578 (vect_analyze_data_ref_dependences): Change max_vf from an int *
18579 to an unsigned int *.
18580 (vect_analyze_data_refs): Change min_vf from an int * to a
18582 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18583 than an unsigned HOST_WIDE_INT.
18584 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
18585 (vect_analyze_data_ref_dependence): Change max_vf from an int *
18586 to an unsigned int *.
18587 (vect_analyze_data_ref_dependences): Likewise.
18588 (vect_compute_data_ref_alignment): Handle polynomial vf.
18589 (vect_enhance_data_refs_alignment): Likewise.
18590 (vect_prune_runtime_alias_test_list): Likewise.
18591 (vect_shift_permute_load_chain): Likewise.
18592 (vect_supportable_dr_alignment): Likewise.
18593 (dependence_distance_ge_vf): Take the vectorization factor as a
18594 poly_uint64 rather than an unsigned HOST_WIDE_INT.
18595 (vect_analyze_data_refs): Change min_vf from an int * to a
18597 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
18598 vfm1 as a poly_uint64 rather than an int. Make the same change
18599 for the returned bound_scalar.
18600 (vect_gen_vector_loop_niters): Handle polynomial vf.
18601 (vect_do_peeling): Likewise. Update call to
18602 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
18603 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
18605 * tree-vect-loop.c (vect_determine_vectorization_factor)
18606 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
18607 (vect_get_known_peeling_cost): Likewise.
18608 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
18609 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
18610 (vect_transform_loop): Likewise. Use the lowest possible VF when
18611 updating the upper bounds of the loop.
18612 (vect_min_worthwhile_factor): Make static. Return an unsigned int
18613 rather than an int.
18614 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
18615 polynomial unroll factors.
18616 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
18617 (vect_make_slp_decision): Likewise.
18618 (vect_supported_load_permutation_p): Likewise, and polynomial
18620 (vect_analyze_slp_cost): Handle polynomial vf.
18621 (vect_slp_analyze_node_operations): Likewise.
18622 (vect_slp_analyze_bb_1): Likewise.
18623 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
18624 than an unsigned HOST_WIDE_INT.
18625 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
18626 (vectorizable_load): Handle polynomial vf.
18627 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
18629 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
18631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18632 Alan Hayward <alan.hayward@arm.com>
18633 David Sherwood <david.sherwood@arm.com>
18635 * match.pd: Handle bit operations involving three constants
18636 and try to fold one pair.
18638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18640 * tree-vect-loop-manip.c: Include gimple-fold.h.
18641 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
18642 niters_maybe_zero parameters. Handle other cases besides a step of 1.
18643 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
18644 Add a path that uses a step of VF instead of 1, but disable it
18646 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
18647 and niters_no_overflow parameters. Update calls to
18648 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
18649 Create a new SSA name if the latter choses to use a ste other
18650 than zero, and return it via niters_vector_mult_vf_var.
18651 * tree-vect-loop.c (vect_transform_loop): Update calls to
18652 vect_do_peeling, vect_gen_vector_loop_niters and
18653 slpeel_make_loop_iterate_ntimes.
18654 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
18655 (vect_gen_vector_loop_niters): Update declarations after above changes.
18657 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
18659 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
18660 128-bit round to integer instructions.
18661 (ceil<mode>2): Likewise.
18662 (btrunc<mode>2): Likewise.
18663 (round<mode>2): Likewise.
18665 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18667 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
18668 unaligned VSX load/store on P8/P9.
18669 (expand_block_clear): Allow the use of unaligned VSX
18670 load/store on P8/P9.
18672 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18674 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
18676 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
18677 swap associated with both a load and a store.
18679 2018-01-02 Andrew Waterman <andrew@sifive.com>
18681 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
18682 * config/riscv/riscv.md (clear_cache): Use it.
18684 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
18686 * web.c: Remove out-of-date comment.
18688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18690 * expr.c (fixup_args_size_notes): Check that any existing
18691 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
18692 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
18693 (emit_single_push_insn): ...here.
18695 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18697 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
18698 (const_vector_encoded_nelts): New function.
18699 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
18700 (const_vector_int_elt, const_vector_elt): Declare.
18701 * emit-rtl.c (const_vector_int_elt_1): New function.
18702 (const_vector_elt): Likewise.
18703 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
18704 of CONST_VECTOR_ELT.
18706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18708 * expr.c: Include rtx-vector-builder.h.
18709 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
18710 directly on the tree encoding.
18711 (const_vector_from_tree): Likewise.
18712 * optabs.c: Include rtx-vector-builder.h.
18713 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
18714 sequence of "u" values.
18715 * vec-perm-indices.c: Include rtx-vector-builder.h.
18716 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
18717 directly on the vec_perm_indices encoding.
18719 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18721 * doc/rtl.texi (const_vector): Describe new encoding scheme.
18722 * Makefile.in (OBJS): Add rtx-vector-builder.o.
18723 * rtx-vector-builder.h: New file.
18724 * rtx-vector-builder.c: Likewise.
18725 * rtl.h (rtx_def::u2): Add a const_vector field.
18726 (CONST_VECTOR_NPATTERNS): New macro.
18727 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
18728 (CONST_VECTOR_DUPLICATE_P): Likewise.
18729 (CONST_VECTOR_STEPPED_P): Likewise.
18730 (CONST_VECTOR_ENCODED_ELT): Likewise.
18731 (const_vec_duplicate_p): Check for a duplicated vector encoding.
18732 (unwrap_const_vec_duplicate): Likewise.
18733 (const_vec_series_p): Check for a non-duplicated vector encoding.
18734 Say that the function only returns true for integer vectors.
18735 * emit-rtl.c: Include rtx-vector-builder.h.
18736 (gen_const_vec_duplicate_1): Delete.
18737 (gen_const_vector): Call gen_const_vec_duplicate instead of
18738 gen_const_vec_duplicate_1.
18739 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
18740 (gen_const_vec_duplicate): Use rtx_vector_builder.
18741 (gen_const_vec_series): Likewise.
18742 (gen_rtx_CONST_VECTOR): Likewise.
18743 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
18744 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18745 Build a new vector rather than modifying a CONST_VECTOR in-place.
18746 (handle_special_swappables): Update call accordingly.
18747 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
18748 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
18749 Build a new vector rather than modifying a CONST_VECTOR in-place.
18750 (handle_special_swappables): Update call accordingly.
18752 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18754 * simplify-rtx.c (simplify_const_binary_operation): Use
18755 CONST_VECTOR_ELT instead of XVECEXP.
18757 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18759 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
18760 the selector elements to be different from the data elements
18761 if the selector is a VECTOR_CST.
18762 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
18763 ssizetype for the selector.
18765 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18767 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
18768 before testing each element individually.
18769 * tree-vect-generic.c (lower_vec_perm): Likewise.
18771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18773 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
18774 * selftest-run-tests.c (selftest::run_tests): Call it.
18775 * vector-builder.h (vector_builder::operator ==): New function.
18776 (vector_builder::operator !=): Likewise.
18777 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
18778 (vec_perm_indices::all_from_input_p): New function.
18779 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
18780 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
18781 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
18782 instead of reading the VECTOR_CST directly. Detect whether both
18783 vector inputs are the same before constructing the vec_perm_indices,
18784 and update the number of inputs argument accordingly. Use the
18785 utility functions added above. Only construct sel2 if we need to.
18787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18789 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
18790 the broadcast of the low byte.
18791 (expand_mult_highpart): Use an explicit encoding for the permutes.
18792 * optabs-query.c (can_mult_highpart_p): Likewise.
18793 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
18794 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18795 (vectorizable_bswap): Likewise.
18796 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
18797 explicit encoding for the power-of-2 permutes.
18798 (vect_permute_store_chain): Likewise.
18799 (vect_grouped_load_supported): Likewise.
18800 (vect_permute_load_chain): Likewise.
18802 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18804 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
18805 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
18806 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
18807 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
18808 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
18809 (vect_gen_perm_mask_any): Likewise.
18811 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18813 * int-vector-builder.h: New file.
18814 * vec-perm-indices.h: Include int-vector-builder.h.
18815 (vec_perm_indices): Redefine as an int_vector_builder.
18816 (auto_vec_perm_indices): Delete.
18817 (vec_perm_builder): Redefine as a stand-alone class.
18818 (vec_perm_indices::vec_perm_indices): New function.
18819 (vec_perm_indices::clamp): Likewise.
18820 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
18821 (vec_perm_indices::new_vector): New function.
18822 (vec_perm_indices::new_expanded_vector): Update for new
18823 vec_perm_indices class.
18824 (vec_perm_indices::rotate_inputs): New function.
18825 (vec_perm_indices::all_in_range_p): Operate directly on the
18826 encoded form, without computing elided elements.
18827 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
18828 encoding. Update for new vec_perm_indices class.
18829 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
18830 the given vec_perm_builder.
18831 (expand_vec_perm_var): Update vec_perm_builder constructor.
18832 (expand_mult_highpart): Use vec_perm_builder instead of
18833 auto_vec_perm_indices.
18834 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
18835 vec_perm_indices instead of auto_vec_perm_indices. Use a single
18836 or double series encoding as appropriate.
18837 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
18838 vec_perm_indices instead of auto_vec_perm_indices.
18839 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18840 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18841 (vect_permute_store_chain): Likewise.
18842 (vect_grouped_load_supported): Likewise.
18843 (vect_permute_load_chain): Likewise.
18844 (vect_shift_permute_load_chain): Likewise.
18845 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
18846 (vect_transform_slp_perm_load): Likewise.
18847 (vect_schedule_slp_instance): Likewise.
18848 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18849 (vectorizable_mask_load_store): Likewise.
18850 (vectorizable_bswap): Likewise.
18851 (vectorizable_store): Likewise.
18852 (vectorizable_load): Likewise.
18853 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
18854 vec_perm_indices instead of auto_vec_perm_indices. Use
18855 tree_to_vec_perm_builder to read the vector from a tree.
18856 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
18857 vec_perm_builder instead of a vec_perm_indices.
18858 (have_whole_vector_shift): Use vec_perm_builder and
18859 vec_perm_indices instead of auto_vec_perm_indices. Leave the
18860 truncation to calc_vec_perm_mask_for_shift.
18861 (vect_create_epilog_for_reduction): Likewise.
18862 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
18863 from auto_vec_perm_indices to vec_perm_indices.
18864 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18865 instead of changing individual elements.
18866 (aarch64_vectorize_vec_perm_const): Use new_vector to install
18867 the vector in d.perm.
18868 * config/arm/arm.c (expand_vec_perm_d::perm): Change
18869 from auto_vec_perm_indices to vec_perm_indices.
18870 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
18871 instead of changing individual elements.
18872 (arm_vectorize_vec_perm_const): Use new_vector to install
18873 the vector in d.perm.
18874 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
18875 Update vec_perm_builder constructor.
18876 (rs6000_expand_interleave): Likewise.
18877 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
18878 (rs6000_expand_interleave): Likewise.
18880 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18882 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
18883 to qimode could truncate the indices.
18884 * optabs.c (expand_vec_perm_var): Likewise.
18886 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
18888 * Makefile.in (OBJS): Add vec-perm-indices.o.
18889 * vec-perm-indices.h: New file.
18890 * vec-perm-indices.c: Likewise.
18891 * target.h (vec_perm_indices): Replace with a forward class
18893 (auto_vec_perm_indices): Move to vec-perm-indices.h.
18894 * optabs.h: Include vec-perm-indices.h.
18895 (expand_vec_perm): Delete.
18896 (selector_fits_mode_p, expand_vec_perm_var): Declare.
18897 (expand_vec_perm_const): Declare.
18898 * target.def (vec_perm_const_ok): Replace with...
18899 (vec_perm_const): ...this new hook.
18900 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
18901 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
18902 * doc/tm.texi: Regenerate.
18903 * optabs.def (vec_perm_const): Delete.
18904 * doc/md.texi (vec_perm_const): Likewise.
18905 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
18906 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
18907 expand_vec_perm for constant permutation vectors. Assert that
18908 the mode of variable permutation vectors is the integer equivalent
18909 of the mode that is being permuted.
18910 * optabs-query.h (selector_fits_mode_p): Declare.
18911 * optabs-query.c: Include vec-perm-indices.h.
18912 (selector_fits_mode_p): New function.
18913 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
18914 is defined, instead of checking whether the vec_perm_const_optab
18915 exists. Use targetm.vectorize.vec_perm_const instead of
18916 targetm.vectorize.vec_perm_const_ok. Check whether the indices
18917 fit in the vector mode before using a variable permute.
18918 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
18919 vec_perm_indices instead of an rtx.
18920 (expand_vec_perm): Replace with...
18921 (expand_vec_perm_const): ...this new function. Take the selector
18922 as a vec_perm_indices rather than an rtx. Also take the mode of
18923 the selector. Update call to shift_amt_for_vec_perm_mask.
18924 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
18925 Use vec_perm_indices::new_expanded_vector to expand the original
18926 selector into bytes. Check whether the indices fit in the vector
18927 mode before using a variable permute.
18928 (expand_vec_perm_var): Make global.
18929 (expand_mult_highpart): Use expand_vec_perm_const.
18930 * fold-const.c: Includes vec-perm-indices.h.
18931 * tree-ssa-forwprop.c: Likewise.
18932 * tree-vect-data-refs.c: Likewise.
18933 * tree-vect-generic.c: Likewise.
18934 * tree-vect-loop.c: Likewise.
18935 * tree-vect-slp.c: Likewise.
18936 * tree-vect-stmts.c: Likewise.
18937 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
18939 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
18940 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
18941 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
18942 (aarch64_vectorize_vec_perm_const): ...this new function.
18943 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18944 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18945 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
18946 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
18947 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18948 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18949 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
18951 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
18952 check for NEON modes.
18953 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
18954 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
18955 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
18956 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
18958 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
18959 the old VEC_PERM_CONST conditions.
18960 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
18961 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
18962 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
18963 (ia64_vectorize_vec_perm_const_ok): Merge into...
18964 (ia64_vectorize_vec_perm_const): ...this new function.
18965 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
18966 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
18967 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
18968 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
18969 * config/mips/mips.c (mips_expand_vec_perm_const)
18970 (mips_vectorize_vec_perm_const_ok): Merge into...
18971 (mips_vectorize_vec_perm_const): ...this new function.
18972 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
18973 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
18974 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
18975 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
18976 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
18977 (rs6000_expand_vec_perm_const): Delete.
18978 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
18980 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18981 (altivec_expand_vec_perm_const_le): Take each operand individually.
18982 Operate on constant selectors rather than rtxes.
18983 (altivec_expand_vec_perm_const): Likewise. Update call to
18984 altivec_expand_vec_perm_const_le.
18985 (rs6000_expand_vec_perm_const): Delete.
18986 (rs6000_vectorize_vec_perm_const_ok): Delete.
18987 (rs6000_vectorize_vec_perm_const): New function.
18988 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
18989 an element count and rtx array.
18990 (rs6000_expand_extract_even): Update call accordingly.
18991 (rs6000_expand_interleave): Likewise.
18992 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
18993 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
18994 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
18995 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
18996 (rs6000_expand_vec_perm_const): Delete.
18997 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
18998 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
18999 (altivec_expand_vec_perm_const_le): Take each operand individually.
19000 Operate on constant selectors rather than rtxes.
19001 (altivec_expand_vec_perm_const): Likewise. Update call to
19002 altivec_expand_vec_perm_const_le.
19003 (rs6000_expand_vec_perm_const): Delete.
19004 (rs6000_vectorize_vec_perm_const_ok): Delete.
19005 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19006 reference to the SPE evmerge intructions.
19007 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19008 an element count and rtx array.
19009 (rs6000_expand_extract_even): Update call accordingly.
19010 (rs6000_expand_interleave): Likewise.
19011 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19012 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19014 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19018 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19019 vector mode and that that mode matches the mode of the data
19021 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19022 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19023 directly using expand_vec_perm_1 when forcing selectors into
19025 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19029 * optabs-query.h (can_vec_perm_p): Delete.
19030 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19031 * optabs-query.c (can_vec_perm_p): Split into...
19032 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19033 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19034 particular selector is valid.
19035 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19036 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19037 (vect_grouped_load_supported): Likewise.
19038 (vect_shift_permute_load_chain): Likewise.
19039 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19040 (vect_transform_slp_perm_load): Likewise.
19041 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19042 (vectorizable_bswap): Likewise.
19043 (vect_gen_perm_mask_checked): Likewise.
19044 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19045 implementations of variable permutation vectors into account
19046 when deciding which selector to use.
19047 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19048 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19049 with a false third argument.
19050 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19051 to test whether the constant selector is valid and can_vec_perm_var_p
19052 to test whether a variable selector is valid.
19054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19056 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19057 * optabs-query.c (can_vec_perm_p): Likewise.
19058 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19059 instead of vec_perm_indices.
19060 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19061 (vect_gen_perm_mask_checked): Likewise,
19062 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
19063 (vect_gen_perm_mask_checked): Likewise,
19065 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19067 * optabs-query.h (qimode_for_vec_perm): Declare.
19068 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
19069 (qimode_for_vec_perm): ...this new function.
19070 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
19072 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19074 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
19075 does not have a conditional at the top.
19077 2018-01-02 Richard Biener <rguenther@suse.de>
19079 * ipa-inline.c (big_speedup_p): Fix expression.
19081 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19084 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
19087 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
19091 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
19092 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
19093 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
19094 cond_taken_branch_cost 3->4.
19096 2018-01-01 Jakub Jelinek <jakub@redhat.com>
19098 PR tree-optimization/83581
19099 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
19100 TODO_cleanup_cfg if any changes have been made.
19102 PR middle-end/83608
19103 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
19104 convert_modes if target mode has the right side, but different mode
19107 PR middle-end/83609
19108 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
19109 last argument when extracting from CONCAT. If either from_real or
19110 from_imag is NULL, use expansion through memory. If result is not
19111 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
19112 the parts directly to inner mode, if even that fails, use expansion
19115 PR middle-end/83623
19116 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
19117 check for bswap in mode rather than HImode and use that in expand_unop
19120 Copyright (C) 2018 Free Software Foundation, Inc.
19122 Copying and distribution of this file, with or without modification,
19123 are permitted in any medium without royalty provided the copyright
19124 notice and this notice are preserved.